AI可视化交互

As the levels of automation and reliance on modern artificial intelligence (AI) approaches increase across multiple industries, the importance of the human-centered perspective becomes more evident. Various actors in such industrial applications, including equipment operators and decision makers, have their needs and preferences that often do not align with the decisions produced by black-box models, potentially leading to mistrust and wasted productivity gain opportunities. In this paper, we examine these issues through the lenses of visual analytics and, more broadly, interactive visualization, and we argue that the methods and techniques from these fields can lead to advances in both academic research and industrial innovations concerning the explainability of AI models. To address the existing gap within and across the research and application fields, we propose a conceptual framework for visual analytics design and evaluation for such scenarios, followed by a preliminary roadmap and call to action for the respective communities.

Trust plays a fundamental role in shaping the willingness of users to engage and collaborate with artificial intelligence (AI) systems. Yet, measuring user trust remains challenging due to its complex and dynamic nature. While traditional survey methods provide trust levels for long conversations, they fail to capture its dynamic evolution during ongoing interactions. Here, we present VizTrust1, which addresses this challenge by introducing a real-time visual analytics tool that leverages a multi-agent collaboration system to capture and analyze user trust dynamics in human-agent communication. Built on established human-computer trust scales—competence, integrity, benevolence, and predictability—, VizTrust enables stakeholders to observe trust formation as it happens, identify patterns in trust development, and pinpoint specific interaction elements that influence trust. Our tool offers actionable insights into human-agent trust formation and evolution in real time through a dashboard, supporting the design of adaptive conversational agents that responds effectively to user trust signals.

Clinical decision-making is often complex and time-consuming due to the large amount of required data, contributing to an increase in mortality rates, particularly in non-cardiac surgery cases. To address this problem, we present an interactive artificial intelligence (AI) dashboard that combines visual analytics (VA) and machine learning (ML) to facilitate quick decision-making and enhance patient care. Visual analytics makes use of human perceptual and cognitive capabilities to quickly process complex data for decision-making. Machine learning is used to predict patients' states for quick decision-making. Though a lot of studies have emerged focusing on the application of VA for patient health care, such as diabetes and infectious diseases, little is known about its application to non-cardiac surgery. In this paper, we harness the capabilities of VA and ML to develop an interactive intelligent dashboard to enhance decision-making for non-cardiac surgery patients. This paper presents the design, development, and initial results of assessing the usability and usefulness of the dashboard with HCI experts and medical doctors. The results showed that users found the dashboard usable and useful. The qualitative analysis revealed six key themes, including the system's role in improving healthcare navigation and equity, its impact on patient-centered care delivery, and the ethical implications of predictive analytics in healthcare. We present our findings along with study limitations and future research directions.

Fashion e-commerce design requires the integration of creativity, functionality, and responsiveness to user preferences. While AI offers valuable support, generative models often miss the nuances of user experience, and task-specific models, although more accurate, lack transparency and real-world adaptability—especially with complex multimodal data. These issues reduce designers’ trust and hinder effective AI integration. To address this, we present FashionCook, a visual analytics system designed to support human–AI collaboration in the context of fashion e-commerce. The system bridges communication among model builders, designers, and marketers by providing transparent model interpretations, “what-if” scenario exploration, and iterative feedback mechanisms. We validate the system through two real-world case studies and a user study, demonstrating how FashionCook enhances collaborative workflows and improves design outcomes in data-driven fashion e-commerce environments.

Efficient and sustainable maritime transport increasingly depends on reliable forecasting and adaptive routing, yet operational adoption remains difficult due to forecast latencies and the need for human judgment in rapid decision-making under changing ocean conditions. We introduce SWR-Viz, an AI-assisted visual analytics framework that combines a physics-informed Fourier Neural Operator wave forecast model with SIMROUTE-based routing and interactive emissions analytics. The framework generates near-term forecasts directly from current conditions, supports data assimilation with sparse observations, and enables rapid exploration of what-if routing scenarios. We evaluate the forecast models and SWR-Viz framework along key shipping corridors in the Japan Coast and Gulf of Mexico, showing both improved forecast stability and realistic routing outcomes comparable to ground-truth reanalysis wave products. Expert feedback highlights the usability of SWR-Viz, its ability to isolate voyage segments with high emission reduction potential, and its value as a practical decision-support system. More broadly, this work illustrates how lightweight AI forecasting can be integrated with interactive visual analytics to support human-centered decision-making in complex geospatial and environmental domains.

With the growing availability of urban data and the increasing complexity of societal challenges, visual analytics has become essential for deriving insights into pressing real-world problems. However, analyzing such data is inherently complex and iterative, requiring expertise across multiple domains. The need to manage diverse datasets, distill intricate workflows, and integrate various analytical methods presents a high barrier to entry, especially for researchers and urban experts who lack proficiency in data management, machine learning, and visualization. Advancements in large language models offer a promising solution to lower the barriers to the construction of analytics systems by enabling users to specify intent rather than define precise computational operations. However, this shift from explicit operations to intent-based interaction introduces challenges in ensuring alignment throughout the design and development process. Without proper mechanisms, gaps can emerge between user intent, system behavior, and analytical outcomes. To address these challenges, we propose Urbanite, a framework for human-AI collaboration in urban visual analytics. Urbanite leverages a dataflow-based model that allows users to specify intent at multiple scopes, enabling interactive alignment across the specification, process, and evaluation stages of urban analytics. Based on findings from a survey to uncover challenges, Urbanite incorporates features to facilitate explainability, multi-resolution definition of tasks across dataflows, nodes, and parameters, while supporting the provenance of interactions. We demonstrate Urbanite's effectiveness through usage scenarios created in collaboration with urban experts. Urbanite is available at urbantk.org/urbanite.

The increasing integration of artificial intelligence (AI) in visual analytics (VA) tools raises vital questions about the behavior of users, their trust, and the potential of induced biases when provided with guidance during data exploration. We present an experiment where participants engaged in a visual data exploration task while receiving intelligent suggestions supplemented with four different transparency levels. We also modulated the difficulty of the task (easy or hard) to simulate a more tedious scenario for the analyst. Our results indicate that participants were more inclined to accept suggestions when completing a more difficult task despite the ai's lower suggestion accuracy. Moreover, the levels of transparency tested in this study did not significantly affect suggestion usage or subjective trust ratings of the participants. Additionally, we observed that participants who utilized suggestions throughout the task explored a greater quantity and diversity of data points. We discuss these findings and the implications of this research for improving the design and effectiveness of ai‐guided va tools.

No abstract available

As social media and online content sharing have grown in popularity, people frequently share documents, photos, and links without being aware of the hidden cybersecurity risks. Current tools typically only check one kind of content at a time and don't display the potential repercussions of risky behaviour. The literature on current cybersecurity techniques is reviewed in this paper, along with their advantages and disadvantages, including the absence of multi-content analysis, poor clarification, and a lack of visual risk reporting. It also draws attention to the problem statement, which calls for a solution that can assess various kinds of content and assist users in analyzing possible outcomes.

This paper presents a method for visualization analysis that utilizes AI to enhance the segmentation of instructional videos into specific topics and to automatically detect deviations in student annotations by assessing both topical relevance and content accuracy. The results are then visualized to help teachers efficiently filter and provide targeted feedback based on identified discrepancies. Collecting sufficient active learning data, such as notes and questions, in online courses is inherently challenging. In our previous work, we developed a video annotation system that enables students to mark specific video frames while adding notes, questions, or comments, thereby lowering the barriers to asking questions. This system successfully gathered a substantial amount of learner-generated annotations, offering valuable insights into cognitive processes. However, the resulting brief, code-mixed annotations pose considerable challenges for instructors to review manually. Traditional Natural Language Processing (NLP) methods struggle to analyze these annotations effectively due to their sparse text, embedded code snippets, and context-dependent meanings. Our approach utilizes LLMs to tackle these analytical difficulties. Evaluation using 359 annotations from 26 students watching a 92-minute Python tutorial demonstrates that our system effectively segmented the video into 13 coherent topics and identified deviations in 20.6% of the annotations. The visualization interface significantly reduced the time instructors spent on reviews by eliminating the need for manual checks, while also providing AI-generated explanations for each identified deviation. This work facilitates data-driven improvements in instruction within large-scale online programming courses, where comprehensive manual review is impractical.

The increasing sophistication in AI models makes Explainable AI (XAI) extremely relevant for enhancing predictions with explanation capabilities. Recent XAI approaches for tabular, image, and graph data have become popular and widely adopted. However, the multi-dimensional nature of spatio-temporal data collected in sensor networks makes most approaches ineffective in such scenarios. XAI approaches that specifically deal with this type of data are emerging, but the complexity of the extracted explanations leads to multi-axis information for nodes, features, and timesteps that is cumbersome to visualize. Indeed, providing effective visualizations could help bridging the existing gap between XAI model predictions and their fruitful and responsible exploitation in practical domains. In this paper, we address this gap by studying the effectiveness of different visualization techniques with multi-dimensional explanations. We adopt a meta-learning XAI framework that identifies salient factors from multiple analytical views. Then, we present a qualitative and quantitative study comparing the effectiveness of 14 visualization techniques using two real-world sensor network datasets. Our results reveal useful patterns, as well as the merits and pitfalls of each technique, paving the way for future work on this topic.

Detecting and analyzing complex patterns in multivariate time-series data is crucial for decision-making in urban and environmental system operations. However, challenges arise from the high dimensionality, intricate complexity, and interconnected nature of complex patterns, which hinder the understanding of their underlying physical processes. Existing AI methods often face limitations in interpretability, computational efficiency, and scalability, reducing their applicability in real-world scenarios. This paper proposes a novel visual analytics framework that integrates two generative AI models, Temporal Fusion Transformer (TFT) and Variational Autoencoders (VAEs), to reduce complex patterns into lower-dimensional latent spaces and visualize them in 2D using dimensionality reduction techniques such as PCA, t-SNE, and UMAP with DBSCAN. These visualizations, presented through coordinated and interactive views and tailored glyphs, enable intuitive exploration of complex multivariate temporal patterns, identifying patterns’ similarities and uncover their potential correlations for a better interpretability of the AI outputs. The framework is demonstrated through a case study on power grid signal data, where it identifies multi-label grid event signatures, including faults and anomalies with diverse root causes. Additionally, novel metrics and visualizations are introduced to validate the models and assess the performance, efficiency, and consistency of latent maps generated by VAE, which have been utilized in prior studies for latent space cartography and used as a benchmark in this study, and the emerging TFT architecture under various configurations. These analyses provide actionable insights for model parameter tuning and reliability improvements. Comparative results highlight that TFT achieves shorter run times and superior scalability to diverse time-series data shapes compared to VAE. This work advances fault diagnosis in multivariate time series, fostering explainable AI to support critical system operations.

Human involvement remains critical in most instances of clinical decision-making. Recent advances in AI and machine learning opened the door for designing, implementing, and translating interactive AI systems to support clinicians in decision-making. Assessing the impact and implications of such systems on patient care and clinical workflows requires in-depth studies. Conducting evaluation studies of AI-supported interactive systems to support decision-making in clinical settings is challenging and time-consuming. These studies involve carefully collecting, analyzing, and interpreting quantitative and qualitative data to assess the performance of the underlying AI-supported system, its impact on the human decision-making process, and the implications for patient care. We have previously developed a toolkit for designing and implementing clinical AI software so that it can be subjected to an application-based evaluation. Here, we present a visual analytics framework for analyzing and interpreting the data collected during such an evaluation process. Our framework supports identifying subgroups of users and patients based on their characteristics, detecting outliers among them, and providing evidence to ensure adherence to regulatory guidelines. We used early-stage clinical AI regulatory guidelines to drive the system design, implemented multiple-factor analysis and hierarchical clustering as exemplary analysis tools, and provided interactive visualizations to explore and interpret results. We demonstrate the effectiveness of our framework through a case study to evaluate a prototype AI-based clinical decision-support system for diagnosing pediatric brain tumors.

In developing machine learning (ML) models for text classification, one common challenge is that the collected data is often not ideally distributed, especially when new classes are introduced in response to changes of data and tasks. In this paper, we present a solution for using visual analytics (VA) to guide the generation of synthetic data using large language models. As VA enables model developers to identify data-related deficiency, data synthesis can be targeted to address such deficiency. We discuss different types of data deficiency, describe different VA techniques for supporting their identification, and demonstrate the effectiveness of targeted data synthesis in improving model accuracy. In addition, we present a software tool, iGAiVA, which maps four groups of ML tasks into four VA views, integrating generative AI and VA into an ML workflow for developing and improving text classification models.

No abstract available

No abstract available

Randomized controlled trials (RCT) are the gold standards for evaluating the efficacy and safety of therapeutic interventions in human subjects. In addition to the pre-specified endpoints, trial participants' experience reveals the time course of the intervention. Few analytical tools exist to summarize and visualize the individual experience of trial participants. Visual analytics allows integrative examination of temporal event patterns of patient experience, thus generating insights for better care decisions. Towards this end, we introduce TrialView, an information system that combines graph artificial intelligence (AI) and visual analytics to enhance the dissemination of trial data. TrialView offers four distinct yet interconnected views: Individual, Cohort, Progression, and Statistics, enabling an interactive exploration of individual and group-level data. The TrialView system is a general-purpose analytical tool for a broad class of clinical trials. The system is powered by graph AI, knowledge-guided clustering, explanatory modeling, and graph-based agglomeration algorithms. We demonstrate the system's effectiveness in analyzing temporal event data through a case study.

No abstract available

Our society increasingly depends on intelligent systems to solve complex problems, ranging from recommender systems suggesting the next movie to watch to AI models assisting in medical diagnoses for hospitalized patients. With the iterative improvement of diagnostic accuracy and efficiency, AI holds significant potential to mitigate medical misdiagnoses by preventing numerous deaths and reducing an economic burden of approximately € 450 billion annually. However, a key obstacle to AI adoption lies in the lack of transparency, that is, many automated systems provide predictions without revealing the underlying processes. This opacity can hinder experts’ ability to trust and rely on AI systems. Visual analytics (VA) provides a compelling solution by combining AI models with interactive visualizations. These specialized charts and graphs empower users to incorporate their domain expertise to refine and improve the models, bridging the gap between AI and human understanding. In this work, the author defines, categorizes, and explores how VA solutions can foster trust across the stages of a typical AI pipeline. The author proposes a design space for innovative visualizations and presents an overview of our previously developed VA dashboards, which support critical tasks within the various pipeline stages, including data processing, feature engineering, hyperparameter tuning, understanding, debugging, refining, and comparing models.

Visual analytics supports data analysis tasks within complex domain problems. However, due to the richness of data types, visual designs, and interaction designs, users need to recall and process a significant amount of information when they visually analyze data. These challenges emphasize the need for more intelligent visual analytics methods. Large language models have demonstrated the ability to interpret various forms of textual data, offering the potential to facilitate intelligent support for visual analytics. We propose LEVA, a framework that uses large language models to enhance users’ VA workflows at multiple stages: onboarding, exploration, and summarization. To support onboarding, we use large language models to interpret visualization designs and view relationships based on system specifications. For exploration, we use large language models to recommend insights based on the analysis of system status and data to facilitate mixed-initiative exploration. For summarization, we present a selective reporting strategy to retrace analysis history through a stream visualization and generate insight reports with the help of large language models. We demonstrate how LEVA can be integrated into existing visual analytics systems. Two usage scenarios and a user study suggest that LEVA effectively aids users in conducting visual analytics.

Understanding collaborative writing dynamics between native speakers (NS) and non-native speakers (NNS) is critical for enhancing collaboration quality and team inclusivity. In this paper, we partnered with communication researchers to develop visual analytics solutions for comparing NS and NNS behaviors in 162 writing sessions across 27 teams. The primary challenges in analyzing writing behaviors are data complexity and the uncertainties introduced by automated methods. In response, we present COALA, a novel visual analytics tool that improves model interpretability by displaying uncertainties in author clusters, generating behavior summaries using large language models, and visualizing writing-related actions at multiple granularities. We validated the effectiveness of COALA through user studies with domain experts (N=2+2) and researchers with relevant experience (N=8). We present the insights discovered by participants using COALA, suggest features for future AI-assisted collaborative writing tools, and discuss the broader implications for analyzing collaborative processes beyond writing.

One prominent field in astronomy seeks to uncover the formation and evolution of the Milky Way based on the large and high-dimensional surveys, thereby deepening our understanding of the Universe. In this context, stellar streams and star clusters are key tracers of the Galaxy’s assembly history. However, they are challenging to isolate in complex, noisy datasets. The Star’s Galactic Origin (StarGO) uses a self-organizing map (SOM) with an adaptive clustering algorithm to identify the galactic origins of halo stars in high-dimensional space. Despite its effectiveness in detecting stellar substructures, the original StarGO results are difficult to interpret due to its clustering process. To address these challenges, we developed an interactive visual analytics system that overcomes this limitation by integrating high-dimensional reduction, adaptive grouping, and multi-view dynamic visualization into a unified interactive workflow. Users can examine spatial, kinematic, and chemical features of halo stars step by step while visually tracing StarGO’s grouping procedure, making the method far more interpretable and transparent. By transforming StarGO into a visual and exploratory platform, our system enhances its scientific usability and greatly improves the interpretability of machine-learning–based stellar population analysis. This interactive extension enables astronomers to more intuitively investigate the origins of stellar streams and clusters, providing a powerful tool for studying the formation and evolution of the Milky Way.

Deep neural networks represent a significant driving force behind the accelerated advancement of artificial intelligence (AI). However, traditional approaches based on manual model architecture design appear inadequate for satisfying various modern applications due to their inherent inefficiency and scalability limitations. The advent of neural architecture search (NAS) methodologies in recent years has rendered the automated search for optimal model architectures a novel possibility. The emergence of NAS benchmarks has also provided an effective solution to the problem of costly evaluation in the NAS process. However, a significant challenge persists that most existing NAS methods cannot provide explanations for the exceptional model architectures identified in the benchmark search space. In light of the vast deployment of visual analytics methods in the domain of explainable AI, this paper introduces a visual analytics framework to facilitate interactive exploration of the search space illustrated in NAS benchmarks. The framework is designed to enable the extraction of potential relationships between model architectures and performance metrics. Furthermore, this paper introduces an analytical pipeline that integrates a range of data processing techniques to efficiently map model architectures across diverse types of large-scale search spaces into a low-dimensional space, which is then adapted for analysis in our visual analytics framework. The effectiveness of the proposed framework is demonstrated through case studies and a user study.

Visual Analytics (VA) integrates humans, data, and models as key actors in insight generation and data-driven decision-making. This position paper values and reflects on 16 VA process models and frameworks and makes nine high-level observations that motivate a fresh perspective on VA. The contribution is the HDMI Canvas, a perspective to VA that complements the strengths of existing VA process models and frameworks. It systematically characterizes diverse roles of humans, data, and models, and how these actors benefit from and contribute to VA processes. The descriptive power of the HDMI Canvas eases the differentiation between a series of VA building blocks, rather than describing general VA principles only. The canvas includes modern human-centered methodologies, including human knowledge externalization and forms of feedback loops, while interpretable and explainable AI highlight model contributions beyond their conventional outputs. The HDMI Canvas has generative power, guiding the design of new VA processes and is optimized for external stakeholders, improving VA outreach, interdisciplinary collaboration, and user-centered design. The utility of the HDMI Canvas is demonstrated through two preliminary case studies.

Urban segregation refers to the physical and social division of people, often driving inequalities within cities and exacerbating socioeconomic and racial tensions. While most studies focus on residential spaces, they often neglect segregation across “activity spaces” where people work, socialize, and engage in leisure. Human mobility data offers new opportunities to analyze broader segregation patterns, encompassing both residential and activity spaces, but challenges existing methods in capturing the complexity and local nuances of urban segregation. This work introduces InclusiViz, a novel visual analytics system for multi-level analysis of urban segregation, facilitating the development of targeted, data-driven interventions. Specifically, we developed a deep learning model to predict mobility patterns across social groups using environmental features, augmented with explainable AI to reveal how these features influence segregation. The system integrates innovative visualizations that allow users to explore segregation patterns from broad overviews to fine-grained detail and evaluate urban planning interventions with real-time feedback. We conducted a quantitative evaluation to validate the model’s accuracy and efficiency. Two case studies and expert interviews with social scientists and urban analysts demonstrated the system’s effectiveness, highlighting its potential to guide urban planning toward more inclusive cities.

This paper looks into integrating customer segmentation based on AI, predictive analytics, and real-time visual analytics to improve marketing decision-making. The research is to leverage deep learning-based clustering and LSTM predictive model to enhance targeted marketing strategies as well as optimize customer engagement and conversion rates. An Autoencoder-based segmentation part was used in combination with supervised learning models (Random Forest, GBM, LSTM) and interactive visual dashboards. The datasets applied in the study are real-world consumer datasets, upon which the model is evaluated based on ARI, Silhouette Score, RMSE, and R² metrics. The segmentation accuracy of the Autoencoder is better than that of K-Means and GMM: an ARI of 0.91. LSTM model proved to have R² of 0.95, which provided significant improvement in the predictive marketing efficiency. This increased response rate by 48% and quadrupled acquisition cost savings by 37%. This validates AI-driven frameworks as superior to all manual segmentation and forecasting methods. With this, the study also shows how AI can be used to optimize marketing efficiency, scalability, precision, and adaptability. This is something that future research needs to consider to further enhance personalized marketing.

Predicting multi-drug use across substances is challenging due to class imbalance and limited interpretability in machine learning models. We propose a four-stage visual analytics workflow that integrates explainable AI and external validation, applied to the UCI drug dataset with seven personality traits. Logistic regression and random forest models achieved high predictive accuracy of 0.98 on frequent labels, while SHAP explanations highlighted Sensation Seeking as the most consistent risk factor, Conscientiousness as protective, and Openness often increasing risk. External validation with mortality and harm signals confirmed consistency, making the framework interpretable, auditable, and actionable for public health and addiction research.

Intelligence analysts perform sensemaking over collections of documents using various visual and analytic techniques to gain insights from large amounts of text. As data scales grow, our work explores how to leverage two AI technologies, large language models (LLMs) and knowledge graphs (KGs), in a visual text analysis tool, enhancing sensemaking and helping analysts keep pace. Collaborating with intelligence community experts, we developed a visual analytics system called VisPile. VisPile integrates an LLM and a KG into various UI functions that assist analysts in grouping documents into piles, performing sensemaking tasks like summarization and relationship mapping on piles, and validating LLM- and KG-generated evidence. Our paper describes the tool, as well as feedback received from six professional intelligence analysts that used VisPile to analyze a text document corpus.

: Monitoring the quality of sleep in patients of sleep disorders is often a time-consuming process, where the clinician manually navigates through large volumes of recorded polysomnography data in an effort to visually detect sleep patterns, such as sleep spindles, sleep stages and hints of disorders. We propose an application that provides healthcare professionals with advanced tools for sleep analysis and spindle detection through visual analytics for pattern detection, AI-based sleep scoring, and an interactive user interface. The system processes multiple physiological signals and provides both raw data visualization, advanced feature analysis capabilities, and a two-dimensional embedding of sleep intervals. By combining signal processing, spindle detection, sleep stage identification and interactive visualization tools, this work helps researchers to efficiently identify, validate, and analyze sleep and spindle characteristics with higher precision than traditional methods.

Intelligent agriculture can renovate agricultural production and management, making agricultural production truly scientific and efficient. The existing data mining technology for agricultural information is powerful and professional. But the technology is not well adapted for intelligent agriculture. Therefore, this paper introduces data visualization and big data analysis into the application scenarios of intelligent agriculture. Firstly, an intelligent agriculture data visualization system was established, and the RadViz data visualization method was detailed for intelligent agriculture. Moreover, the intelligent agriculture data were processed using dimensionality reduction through principal component analysis (PCA) and further optimized through k-means clustering (KMC). Finally, the crop yield was predicted using the multiple regression algorithm and the residual principal component regression algorithm. The crop yield prediction model was proved effective through experiments.

This paper proposes a novel recognition approach for operation states of hydroelectric generating units based on data fusion and visualization analysis. First, the principal component analysis (PCA) is employed to simplify signals from multiple channels into a single fused signal, thereby reducing data computation for multiple‐channel signals. To reflect the features of fused signals under different operation states, the Gramian angular field (GAF) method is applied to convert the fused signals into image formats, including Gramian angular differential field (GADF) images and Gramian angular summation field (GASF) images, then a depthwise separable convolution neural network (DSCNN) model is established to achieve the operation state recognition for the unit by GADF and GASF images. Based on the operation data from a Kaplan hydroelectric unit at a hydropower station in Southwest China, an experiment on operation recognition is conducted. The proposed PCA–GAF–DSCNN method achieves an accuracy rate of 95.21% with GADF images and 96.41% with GASF images, which were higher than the results obtained using original signals with the GAF–DSCNN method. The results indicate that the fused signal with PCA demonstrates superior performance in the operation recognition compared to the original signals, and PCA–GAF–DSCNN can be used for hydroelectric units effectively. This approach accurately identifies abnormal states in units, making it suitable for monitoring and fault diagnosis in the daily operations of hydroelectric generating units.

No abstract available

The rapid expansion of bioinformatics and healthcare data presents both opportunities and challenges for deriving useful information for biomedical research and clinical decision-making. The amount, diversity, and complexity of such data frequently pose challenges for conventional information retrieval techniques. A framework for intelligent information retrieval and visualization enabled by big data that combines machine learning, natural language processing, and semantic technologies is presented in this study. The suggested method effectively processes massive biomedical datasets by utilizing distributed analytics, which improves retrieval precision and scalability. In order to convert multidimensional data into comprehensible patterns that support precision medicine and disease prediction, advanced visualisation techniques are used. When compared to baseline models, experimental validation on real-world healthcare datasets shows significant gains in retrieval accuracy (up to 18%), decreased query latency, and improved interpretability. The results highlight the technological importance of integrating semantic intelligence and big data analytics to further research in bioinformatics and healthcare informatics.

The research puts forward solutions from three levels: in the aspect of heterogeneous data fusion, a semantic alignment model based on dynamic Knowledge Graph (KG) is constructed, and semantic conflicts are solved through multimodal ontology modeling and dynamic entity alignment mechanism (fusion of structure, semantics and temporal similarity); On the level of intelligent analysis, a collaborative hybrid computing framework based on improved Lambda architecture is designed, which combines real-time processing of sliding window and deep mining of batch data lakes, and realizes the optimization of results through dynamic weight fusion. In the visual interaction, a three-tier engine integrating progressive rendering and predictive loading is developed. By using quadtree index, dynamic level of detail (LOD) and LSTM behavior prediction, the interaction delay at 1080P resolution is controlled within 100ms. The experimental results show that the cross-modal link accuracy rate of this technical scheme is 92.7%, the semantic conflict resolution rate is 91.2%, the batch processing delay is stable at 210±15ms, the throughput is increased by 18.75%, and the visual interaction delay is significantly reduced, which achieves a performance breakthrough in theory and application.

With the rapid development of cloud computing and Internet technologies, the scale and complexity of network traffic data have grown exponentially. Traditional operation and maintenance methods face severe challenges in the real-time analysis and monitoring of massive data. This paper designs and implements a cloud server traffic data visualization platform that integrates data collection, storage, visualization, and intelligent analysis. The platform adopts a three-tier architecture: the data collection layer uses VPS probes distributed globally to obtain key performance indicators such as bandwidth, latency, and packet loss rate in real time; the data processing layer performs time-series data aggregation and storage based on Prometheus and MySQL; the visualization layer builds an interactive dashboard supporting multiple chart types and geographic information display by integrating Grafana and ECharts. In addition, the platform innovatively introduces the thought chain reasoning and retrieval-augmented generation technologies of large language models to realize automatic analysis of node health status and report generation. Practical deployment shows that the platform can effectively improve the efficiency of network operation and maintenance and provide intuitive and intelligent decision support for cloud resource management.

No abstract available

Abstract Summary The vast generation of genetic data poses a significant challenge in efficiently uncovering valuable knowledge. Introducing GENEVIC, an AI-driven chat framework that tackles this challenge by bridging the gap between genetic data generation and biomedical knowledge discovery. Leveraging generative AI, notably ChatGPT, it serves as a biologist’s “copilot.” It automates the analysis, retrieval, and visualization of customized domain-specific genetic information, and integrates functionalities to generate protein interaction networks, enrich gene sets, and search scientific literature from PubMed, Google Scholar, and arXiv, making it a comprehensive tool for biomedical research. In its pilot phase, GENEVIC is assessed using a curated database that ranks genetic variants associated with Alzheimer’s disease, schizophrenia, and cognition, based on their effect weights from the Polygenic Score (PGS) Catalog, thus enabling researchers to prioritize genetic variants in complex diseases. GENEVIC’s operation is user-friendly, accessible without any specialized training, secured by Azure OpenAI’s HIPAA-compliant infrastructure, and evaluated for its efficacy through real-time query testing. As a prototype, GENEVIC is set to advance genetic research, enabling informed biomedical decisions. Availability and implementation GENEVIC is publicly accessible at https://genevicanath2024.streamlit.app. The underlying code is open-source and available via GitHub at https://github.com/bsml320/GENEVIC.git (also at https://github.com/anath2110/GENEVIC.git).

No abstract available

This study aims to design a deep data display and interactive operation platform for the education industry, focusing on the needs of smart education scenarios, combine virtual reality technology to improve the interpretation and interactive experience of educational information data, adopt a deep experience technology system based on Unity and Unreal Engine, and use D3.s, Th.s and other graphic tools for data presentation. The system preprocesses, fills in values, and transforms space for educational materials, successfully constructs a three-dimensional display of complex data, and improves students’ participation and learning effects through gesture capture, touch operation and other interactive means. By incorporating optical perception technology into the gesture tracking and interaction modules, the system achieves more precise recognition of hand movements and enhances the stability of immersive operations. By integrating intelligent interaction mechanisms and real-time data feedback, the platform actively supports smart education ecosystems. The system’s effect test report in multiple course areas shows that the teaching effect and learning efficiency have been significantly improved, especially in customized recommendation services and tutor adjustment feedback speed. In terms of user experience, the system’s ease of operation, user interaction performance, and operational stability have all been highly recognized. The conclusion shows that the system has great development prospects in terms of education quality and learning experience.

This study explores how combining the YOLOv10 artificial intelligence model with data visualization technology can improve the speed and accuracy of sorting and delivering packages in logistics. To handle challenges like varying package sizes and environmental noise in real-world scenarios, we enhanced the YOLOv10 system by adding a new module called the Multi-Scale Dynamic Feature Fusion (MDFM). This module works with advanced techniques like full-dimensional dynamic convolution and combines KANConv with C2f components to better recognize different-sized packages. Through experiments, we confirmed these improvements significantly boosted detection accuracy and speed, with key metrics like precision, recall, and $\mathbf{F 1}$-scores showing clear gains. We also integrated data visualization tools to display logistics information more clearly, helping teams spot delays, reduce manual sorting efforts, and cut down errors and wasted resources.By making complex data easier to understand, the system supports faster decision-making and smoother operations. These advancements provide a strong foundation for automating logistics processes and driving the growth of smart, AI-driven delivery systems.

The aim of personalized education is to tailor the learning experience in accordance with individual students’ academic needs and maximize their academic performance through these customized approaches. However, the unavailability of real-time insights and personalized guidance poses a major challenge in the prediction of students’ grades and the development of customized study plans. To overcome these limitations, our research proposes the Student Performance Analysis Tool (SPAT), a machine learning-based analytical platform implemented in Python. SPAT was built using a dataset of 145 university students from U.S. university data. The tool focuses on predicting student performance based on academic and non-academic features which is novel in execution. Further, the tool investigates on seven machine learning models such as Logistic Regression, Decision Tree, Random Forest, Support Vector Machine (SVM), Gradient Boosting, XGBoost, and AdaBoost. Also, SPAT combines real-time data visualization and grade prediction with an interactive interface, allowing teachers and students to make data-driven, informed decisions for academic enhancement. Results represented by SPAT indicates high accuracy for predicting the student rate for algorithms Gradient Boost and XGBoost as 79%. Through the use of AI-powered analytics, the tool proposes adaptive learning methods, showing the power of machine learning in developing data-driven with tailored education.

Panoramic radiography is an essential auxiliary diagnostic tool for oral diseases. It is a difficult and time-consuming task to conduct extensive panoramic radiography interpretation. These challenges are exacerbated by the creation of electronic medical records and the investigation of oral diseases using collective data. So, we develop a visualization system based on panoramic radiographs. Its function focuses on the intelligent diagnosis and statistical analysis of oral diseases. Firstly, we provide a human-machine collaborative tool for the diagnosis and data extraction of oral diseases in panoramic radiographs. After that, the system generates electronic medical records, including visual charts of oral health status and radiology reports. We further develop statistical correlation analysis to visually evaluate and interactively explore the statistical data from oral health surveys. We conduct intelligent diagnosis, obtain the electronic medical records and do collective analysis based on 521 panoramic radiographs. The available analyses cover disease-prone teeth, disease distribution per tooth position and association of age, sex with oral diseases. The results are reported from a comprehensive case study showing that our system can improve the efficiency in disease detection and data mining. It can also fuel research studies in the field of public oral health and provide robust support for oral healthcare strategies.

No abstract available

Focusing on the core role of intelligent AGV systems in smart warehousing, this paper proposes a hybrid navigation architecture integrating magnetic guidance and visual SLAM, combined with visualization technology to achieve full-process monitoring of warehouse operations. The system enhances environmental adaptability through multi-source sensor data fusion (magnetic guidance accuracy: ±1 cm, visual SLAM dynamic correction). Efficient obstacle avoidance is realized via A* global path planning and the Dynamic Window Approach (DWA). A digital twin visualization platform is constructed using the Three.js engine, supporting real-time AGV trajectory rendering and anomaly warnings. The design adopts a distributed fault-tolerant mechanism (hardware redundancy + software degradation) to ensure system reliability, providing scalable technical references for warehouse automation upgrades.

No abstract available

The present modelling aims to construct a computational information representation system useful for decision support system (DSS) solutions in the realization of intelligent systems or complex systems analysis solutions. Starting from an n-dimensional space (with n ≥ 7) represented by problem variables (referred to as CSF—Critical Success Factors), a dimensional embedding procedure is used to transition to a two-dimensional space. In the two-dimensional space, thanks to new lattice motion algorithms, the decision support system can determine the optimal solution with a lower computational cost based on the decision-maker’s preferences. Finally, thanks to an algorithm that takes into account the hierarchical order of importance of the seven CSFs as per the expert’s liking or according to his optimization logics, a return is made to the n-dimensional space and the final solution in the original space. As we will see, the starting and ending states in the n-dimensional space (referred to as micro-states) when projected into the two-dimensional space generate states (referred to as macro-states) which are degenerate. In other words, the correspondence between micro-states and macro-states is not one-to-one, as multiple micro-states correspond to one macro-state. Therefore, in relation to the decision-maker’s preferences, it will be the responsibility of the decision support system to provide the decision-maker with the micro-state of interest in the n-dimensional space (dimensional emergence procedure), starting from the obtained optimal macro-state. This result can be achieved starting from a flat chain of sensors capable of measuring/emulating certain specific parameters of interest. As we will see, it emerges that by considering random–exhaustive rolling value paths in order to track and potentially intervene to rebalance a dynamic system representing the state of stress/sensing of a system of interest, we are using the most general and, therefore, complex hypotheses of ergodic theory. In this work, we will focus on the representation of information in n-dimensional and two-dimensional spaces, as well as construct evaluation scenarios. We will also show the results of the decision support system in some cases of specific interest, thanks to a specific lattice motion algorithm of the realized decision-making environment.

A comprehensive understanding of physiological data is essential for anesthesiologists to monitor and maintain the health of surgical patients. Understanding anesthetic signals is essential to improve over the past ten decades to comprehend a variety of diseases, ensure the safety of patients, and accelerate their recovery. We identified and summarized four themes of artificial intelligence (AI) research in anesthesia: applications of AI in anesthesiology, signals monitored in different anesthesia stages, commonly monitored signs, and data standardization. The effect of various AI technologies on data analysis was studied, and the types, functions, sensors, characteristics, and intelligent analysis of physiological signals monitored during anesthesia are presented in this article. AI influences anesthesia practices in many ways, including preoperative data evaluation, anesthetic depth monitoring, and postoperative event prediction.

No abstract available

Introduction: Gas chromatography combined with mass spectrometry (GC/MS) is popular analytical instrumentation for chemical separation and identification. A novel framework for chemical forensics based on the visualization of GC/MS data and transfer learning is proposed. Methods: To evaluate the framework, 228 GC/MS data collected from two standard cannabis varieties, i.e., hemp and marijuana, were utilized. By processing the raw GC/MS data, analytical features, including retention times, mass-to-charge ratios, intensities, and summed ion mass spectra, were successfully transformed into two types of image representations. The GC/MS data transformed images were fed into a pre-trained convolutional neural network (CNN) to develop intelligent classifiers for the sample classification tasks. The effectiveness of several hyper-parameters for improving classification performance was investigated during transfer learning. Results: The proposed analytical workflow could classify hemp and marijuana with 97% accuracy. Furthermore, the transfer-learning-based classifiers were established without requiring big data sets and peak alignment. Discussion: The potential application of the new artificial intelligence (AI)-powered framework for chemical forensics using GC/MS data has been demonstrated. This framework provides unique opportunities for classifying various types of physical evidence using chromatography and mass spectrometry signals.

This paper introduces an innovative multiactor framework that harnesses the potential of LLMs to augment the functionalities of ICS. By integrating conversational AI technologies, this framework significantly improves human-machine interactions, enabling sophisticated analysis and visualization of intricate data sets. The core of the system comprises specialized LLM actors that interact through a LangGraph-based multiactor framework, addressing various aspects of IEC 61499 control systems including PLC code analysis, SQL query execution, and data visualization. This integration enables operators to interact with the control system using natural language, significantly reducing technical barriers and enhancing the accessibility and usability of complex industrial systems.

With the rapid development of artificial intelligence, data visualization analysis platforms have been widely applied in various fields. This study mainly explores the optimization and performance improvement of algorithms for data visualization analysis platforms based on artificial intelligence. Firstly, the definition of a data visualization analysis platform and the application of artificial intelligence in it were introduced, and the current problems and challenges were pointed out. Then, a discussion was conducted on algorithm optimization for various stages of research, including data preprocessing, data clustering, data classification, and optimization of data association analysis algorithms. Subsequently, the study proposed some performance improvement methods, including the application of parallel computing technology, distributed computing technology, data compression technology, and data indexing technology.

Plant diseases significantly impact crop productivity and quality, posing a serious threat to global agriculture. The process of identifying and categorizing these diseases is often time-consuming and prone to errors. This research addresses this issue by employing a convolutional neural network and support vector machine (CNN-SVM) hybrid model to classify diseases in four economically important crops: strawberries, peaches, cherries, and soybeans. The objective is to categorize 10 classes of diseases, with six diseased classes and four healthy classes, for these crops using the deep learning-based CNN-SVM model. Several pre-trained models, including VGG16, VGG19, DenseNet, Inception, MobileNetV2, MobileNet, Xception, and ShuffleNet, were also trained, achieving accuracy ranges from 53.82% to 98.8%. The proposed model, however, achieved an average accuracy of 99.09%. While the proposed model's accuracy is comparable to that of the VGG16 pre-trained model, its significantly lower number of trainable parameters makes it more efficient and distinctive. This research demonstrates the potential of the CNN-SVM model in enhancing the accuracy and efficiency of plant disease classification. The CNN-SVM model was selected over VGG16 and other models due to its superior performance metrics. The proposed model achieved a 99% F1-score, a 99.98% Area Under the Curve (AUC), and a 99% precision value, demonstrating its efficacy. Additionally, class activation maps were generated using the Gradient Weighted Class Activation Mapping (Grad-CAM) technique to provide a visual explanation of the detected diseases. A heatmap was created to highlight the regions requiring classification, further validating the model's accuracy and interpretability.

Network intrusion detection systems (IDS) are crucial, but cybersecurity professionals have a difficult time trusting and acting on the predictions made by many IDS models based on machine learning owing to the lack of transparency in these models. Conventional models work well for attack detection, but their lack of transparency makes them unsuitable for incident response. This paper presents a novel hybrid approach to intrusion detection (ID). It integrates Random Forest (RF) for classification with Attention-Based Neural Networks (Ab-NNs) for more in-depth insights and interpretability at the feature level. Improved detection accuracy is a result of the attention-based model's ability to detect complex patterns in the data. In contrast, the RF model classifies network traffic as either an attack or not. This study meets the need for being able to explain things by using SHAP (SHapley Additive Explanation) along with LIME (Local Interpretable Model-Agnostic Explanations), which give the model's decisions both global as well as local meanings. Due to these visualisations, cybersecurity professionals could better understand the reasons behind detected attacks. Experimental results on datasets like NSL-KDD and CICIDS show that the proposed approach attains high detection performance (98% accuracy) and provides transparency (local decision reasons, feature importance).

No abstract available

Simple Summary An AI-based approach for diagnosing oral squamous cell carcinoma is developed using a database of histopathology images obtained through biopsy and evaluated by two pathologists. To increase the objectivity and repeatability of the histopathological examination, automated multiclass grading of OSCC is performed in the first step. Furthermore, the second step builds confidence in the AI-based system by integrating explainable AI components like Grad-CAM, which give clinicians valuable visual insights into the model’s decision-making process. Regarding multiclass grading, the method based on deep convolutional neural networks produced satisfactory results.

English grammar mastery continues to be a challenge for language learners, particularly in understanding abstract rules and independently correcting errors. While AIdriven grammar correction tools have achieved strong performance, most lack transparency and interactive, learnercentered feedback. Models such as BERT, RoBERTa, and GECToR offer high accuracy but limited explainability or visual learning support. This study presents a system using Streamlit, Python, and spaCy that accepts user input, detects grammatical errors, and provides interactive visual feedback with grammar rule explanations. The system integrates GPT3.5 for correction and SHAP for token-level explanation, enhancing both accuracy and interpretability. Evaluated on the English Grammatical Error Benchmark dataset, it achieves 92.4 % accuracy, with a BLEU score of 75.8 and ROUGE-L score of 77.3. To our knowledge, this is one of the first systems to combine explainable AI and conceptual visualization for grammar correction in educational contexts, offering both technical performance and pedagogical value.

The study proposes a novel deep learning model, a Convolutional Autoencoder with modified loss functions (CAE-MLS) for explainable breast cancer diagnosis through histopathology image analysis. To enhance model interpretability, we used Gradient-weighted Class Activation Mapping (GradCAM) visualization, which emphasizes areas of interest that affect the model's diagnostic decisions. The optimization of the model is achieved by using a custom loss function which combines mean absolute error, mean squared error, and structural similarity index to balance pixel-level accuracy with quality in image reconstruction. Our model achieves 91\% classification accuracy, with 93% precision for Malignant tumour detection. The proposed model attained a best-balanced performance for both classes, with F1-scores of 0.88 and 0.92 for benign and malignant cases in the 10x magnifications, respectively. Grad-CAM visualizations confirm the model's ability to focus on clinically relevant image regions at both 10x and 40x magnifications, providing healthcare professionals with transparent insights into the decision-making process.

Social media platforms today strive to improve user experience through AI recommendations, yet the value of such recommendations vanishes as users do not understand the reasons behind them. This issue arises because explainability in social media is general and lacks alignment with user-specific needs. In this vision paper, we outline a user-segmented and context-aware explanation layer by proposing a visual explanation system with diverse explanation methods. The proposed system is framed by the variety of user needs and contexts, showing explanations in different visualized forms, including a technically detailed version for AI experts and a simplified one for lay users. Our framework is the first to jointly adapt explanation style (visual vs. numeric) and granularity (expert vs. lay) inside a single pipeline. A public pilot with 30 X users will validate its impact on decision-making and trust.

No abstract available

Explainable AI: Scene Classification and GradCam Visualization focuses on developing deep learning models to predict landscape types in images, particularly satellite images, for real-world applications such as landscape recognition. Through this initiative, participants will probe into the basic theory of Stacked Neural Networks, CNNs, and residual nets to gain a comprehensive understanding of their operation and applications. Using Python libraries, participants will learn image import, pre- processing, and visualization techniques, along with data augmentation techniques to improve model generalization. The core of the project is to use Keras and TensorFlow 2.0 to form a CNN-based model with residual blocks, and then compile and train the model. Evaluation metrics such as precision, precision, and recall are used to calculate model performance and generalization capabilities. Additionally, participants will explore Grad-CAM, a technology from Explainable AI that visualizes the activation maps used by CNNs for predictions. In the conclusion of the project, participants will hone their skills in stacked learning, image processing, and interpretability techniques, and gain concrete insights into how AI models work in landscape classification.

No abstract available

Agriculture plays a crucial role in Bangladesh's economy, with corn serving as a key crop. Plant diseases pose significant threats to agricultural productivity and economic stability worldwide. Effective monitoring and prediction are essential to mitigate these risks, as the prevalence of infectious diseases severely disrupts agricultural production. This study focuses on diagnosing agricultural diseases and pests affecting corn stalks using convolutional neural networks (CNN) combined with an SVM classifier. The proposed model achieves a classification accuracy of 98.34%, comparable to transfer learning models such as VGG-16 (97.42%) and VGG-19 (97.13%), while utilizing approximately 0.95 million parameters—about 145 times and 151 times fewer than VGG-16 and VGG-19, respectively. The model exhibits outstanding performance, with accuracy, recall, and F1 scores nearing 97% and an exceptional Area Under Curve (AUC) score of 99%. Furthermore, the model's compact design requires minimal disk space (approximately 4 MB) and features a significantly reduced parameter count. To enhance interpretability, the study integrates explainable AI techniques, including Gradient Weighted Class Activation Mapping (Grad-CAM), Shapley Additive Explanations (SHAP), and Local Interpretable Model-Agnostic Explanations (LIME). These methods provide visual explanations through heatmaps, highlighting regions critical for classification, thereby facilitating a clearer understanding of the disease diagnosis process.

No abstract available

Incorporating style-related objectives into shape design has been centrally important to maximize product appeal. However, stylistic features such as aesthetics and semantic attributes are hard to codify even for experts. As such, algorithmic style capture and reuse have not fully benefited from automated data-driven methodologies due to the challenging nature of design describability. This paper proposes an AI-driven method to fully automate the discovery of brand-related features. Our approach introduces BIGNet, a two-tier Brand Identification Graph Neural Network (GNN) to classify and analyze scalar vector graphics (SVG). First, to tackle the scarcity of vectorized product images, this research proposes two data acquisition workflows: parametric modeling from small curve-based datasets, and vectorization from large pixel-based datasets. Secondly, this study constructs a novel hierarchical GNN architecture to learn from both SVG’s curve-level and chunk-level parameters. In the first case study, BIGNet not only classifies phone brands but also captures brand-related features across multiple scales, such as the location of the lens, the height-width ratio, and the screen-frame gap, as confirmed by AI evaluation. In the second study, this paper showcases the generalizability of BIGNet learning from a vectorized car image dataset and validates the consistency and robustness of its predictions given four scenarios. The results match the difference commonly observed in luxury vs. economy brands in the automobile market. Finally, this paper also visualizes the activation maps generated from a convolutional neural network and shows BIGNet’s advantage of being a more human-friendly, explainable, and explicit style-capturing agent.

The intelligent transportation system, especially autonomous vehicles, has seen a lot of interest among researchers owing to the tremendous work in modern artificial intelligence (AI) techniques, especially deep neural learning. As a result of increased road accidents over the last few decades, significant industries are moving to design and develop autonomous vehicles. Understanding the surrounding environment is essential for understanding the behavior of nearby vehicles to enable the safe navigation of autonomous vehicles in crowded traffic environments. Several datasets are available for autonomous vehicles focusing only on structured driving environments. To develop an intelligent vehicle that drives in real-world traffic environments, which are unstructured by nature, there should be an availability of a dataset for an autonomous vehicle that focuses on unstructured traffic environments. Indian Driving Lite dataset (IDD-Lite), focused on an unstructured driving environment, was released as an online competition in NCPPRIPG 2019. This study proposed an explainable inception-based U-Net model with Grad-CAM visualization for semantic segmentation that combines an inception-based module as an encoder for automatic extraction of features and passes to a decoder for the reconstruction of the segmentation feature map. The black-box nature of deep neural networks failed to build trust within consumers. Grad-CAM is used to interpret the deep-learning-based inception U-Net model to increase consumer trust. The proposed inception U-net with Grad-CAM model achieves 0.622 intersection over union (IoU) on the Indian Driving Dataset (IDD-Lite), outperforming the state-of-the-art (SOTA) deep neural-network-based segmentation models.

No abstract available

In this work, we have proposed a deep convolutional neural network model with the background of a network as ResNet and a classification block as a faster RCNN model to predict multiple lung cancer abnormalities from X-ray images. Experimentations are conducted on Kaggle image dataset repositories with X-ray images of about 55000 images comprising of 15 different classes. We have proposed a region proposal network capable of autonomously learning from the pre-trained data specifications and applying a bounding box to cancer-affected regions to detect the affected areas. About 80% of samples were considered to conduct training from each class, and 20% were used for testing. The results obtained have an accuracy of about 96% for classification.

Artificial intelligence and deep learning are powerful tools for extracting knowledge from large datasets, particularly in healthcare. However, their black-box nature raises interpretability concerns, especially in high-stakes applications. Existing eXplainable Artificial Intelligence methods often focus solely on visualization or rule-based explanations, limiting interpretability's depth and clarity. This work proposes a novel explainable AI method specifically designed for medical image analysis, integrating statistical, visual, and rule-based explanations to improve transparency in deep learning models. Statistical features are derived from deep features extracted using a custom Mobilenetv2 model. A two-step feature selection method - zero-based filtering with mutual importance selection - ranks and refines these features. Decision tree and RuleFit models are employed to classify data and extract human-readable rules. Additionally, a novel statistical feature map overlay visualization generates heatmap-like representations of three key statistical measures (mean, skewness, and entropy), providing both localized and quantifiable visual explanations of model decisions. The proposed method has been validated on five medical imaging datasets - COVID-19 radiography, ultrasound breast cancer, brain tumor magnetic resonance imaging, lung and colon cancer histopathological, and glaucoma images - with results confirmed by medical experts, demonstrating its effectiveness in enhancing interpretability for medical image classification tasks.

BACKGROUND Accurate and efficient classification of brain tumors, including gliomas, meningiomas, and pituitary adenomas, is critical for early diagnosis and treatment planning. Magnetic resonance imaging (MRI) is a key diagnostic tool, and deep learning models have shown promise in automating tumor classification. However, challenges remain in achieving high accuracy while maintaining interpretability for clinical use. METHODS This study explores the use of transfer learning with pre-trained architectures, including VGG16, DenseNet121, and Inception-ResNet-v2, to classify brain tumors from MRI images. An ensemble-based classifier was developed using a majority voting strategy to improve robustness. To enhance clinical applicability, explainability techniques such as Grad-CAM++ and Integrated Gradients were employed, allowing visualization of model decision-making. RESULTS The ensemble model outperformed individual Convolutional Neural Network (CNN) architectures, achieving an accuracy of 86.17% in distinguishing gliomas, meningiomas, pituitary adenomas, and benign cases. Interpretability techniques provided heatmaps that identified key regions influencing model predictions, aligning with radiological features and enhancing trust in the results. CONCLUSIONS The proposed ensemble-based deep learning framework improves the accuracy and interpretability of brain tumor classification from MRI images. By combining multiple CNN architectures and integrating explainability methods, this approach offers a more reliable and transparent diagnostic tool to support medical professionals in clinical decision-making.

Accurate and interpretable classification of brain tumors from magnetic resonance imaging (MRI) is critical for effective diagnosis and treatment planning. This study presents an ensemble-based deep learning framework that combines MobileNetV2 and DenseNet121 convolutional neural networks (CNNs) using a soft voting strategy to classify three common brain tumor types: glioma, meningioma, and pituitary adenoma. The models were trained and evaluated on the Figshare dataset using a stratified 5-fold cross-validation protocol. To enhance transparency and clinical trust, the framework integrates an Explainable AI (XAI) module employing Grad-CAM++ for class-specific saliency visualization, alongside a symbolic Clinical Decision Rule Overlay (CDRO) that maps predictions to established radiological heuristics. The ensemble classifier achieved superior performance compared to individual CNNs, with an accuracy of 91.7%, precision of 91.9%, recall of 91.7%, and F1-score of 91.6%. Grad-CAM++ visualizations revealed strong spatial alignment between model attention and expert-annotated tumor regions, supported by Dice coefficients up to 0.88 and IoU scores up to 0.78. Clinical rule activation further validated model predictions in cases with distinct morphological features. A human-centered interpretability assessment involving five board-certified radiologists yielded high Likert-scale scores for both explanation usefulness (mean = 4.4) and heatmap-region correspondence (mean = 4.0), reinforcing the framework's clinical relevance. Overall, the proposed approach offers a robust, interpretable, and generalizable solution for automated brain tumor classification, advancing the integration of deep learning into clinical neurodiagnostics.

Chemical sensor arrays mimic the mammalian olfactory system to achieve artificial olfaction, and receptor materials resembling olfactory receptors are being actively developed. To realize practical artificial olfaction, it is essential to provide guidelines for developing effective receptor materials based on the structure–activity relationship. In this study, we demonstrated the visualization of the relationship between sensing signal features and odorant molecular features using an explainable AI (XAI) technique. We focused on classification tasks and employed a convolutional neural network (CNN) and score-class activation mapping (Score-CAM) methods. The results obtained from analyzing the 94 odor samples prepared using pure solvents indicate that the information regarding the active receptor materials and data points in the signals and the structure–activity relationship could be accurately extracted. Therefore, using XAI techniques to analyze sensor signals from odor data is an important technique for advancing artificial olfaction.

In high-stakes decision-making, such as healthcare diagnostics, financial forecasting, and legal judgments, trust and transparency are key in AI systems. LLMs have indeed proved to be very capable in processing and generating human-like text, making them invaluable tools in these critical sectors. However, the opacity of their decision-making processes poses a significant challenge for accountability and ethical compliance. This paper introduces comprehensive Explainable AI frameworks tailored for LLMs with the goal of improving their interpretability and reliability for high-stakes environments. Using publicly available real-world datasets, we systematically study the effectiveness of various XAI techniques, including attention visualization, feature importance mapping, and counterfactual explanations. Our findings point to the fact that such embedding serves not only to demystify the decision pathways of LLMs but also to bring their operations in line with regulatory standards and stakeholder expectations. This research makes its contribution to responsible LLM deployment in situations where mistakes have significant consequences by closing the gap between sophisticated AI capabilities and the need for transparent decision-making. The proposed XAI frameworks will, in the end, be a starting point for trust, ethics, and wider acceptance of AI-driven solutions in high-stakes domains.

Retinal diseases are among the leading causes of blindness worldwide, requiring early detection for effective treatment. Manual interpretation of ophthalmic imaging, such as optical coherence tomography (OCT), is traditionally time-consuming, prone to inconsistencies, and requires specialized expertise in ophthalmology. This study introduces OculusNet, an efficient and explainable deep learning (DL) approach for detecting retinal diseases using OCT images. The proposed method is specifically tailored for complex medical image patterns in OCTs to identify retinal disorders, such as choroidal neovascularization (CNV), diabetic macular edema (DME), and age-related macular degeneration characterized by drusen. The model benefits from Saliency Map visualization, an Explainable AI (XAI) technique, to interpret and explain how it reaches conclusions when identifying retinal disorders. Furthermore, the proposed model is deployed on a web page, allowing users to upload retinal OCT images and receive instant detection results. This deployment demonstrates significant potential for integration into ophthalmic departments, enhancing diagnostic accuracy and efficiency. In addition, to ensure an equitable comparison, a transfer learning approach has been applied to four pre-trained models: VGG19, MobileNetV2, VGG16, and DenseNet-121. Extensive evaluation reveals that the proposed OculusNet model achieves a test accuracy of 95.48% and a validation accuracy of 98.59%, outperforming all other models in comparison. Moreover, to assess the proposed model's reliability and generalizability, the Matthews Correlation Coefficient and Cohen's Kappa Coefficient have been computed, validating that the model can be applied in practical clinical settings to unseen data.

Spin coating polymer thin films to achieve specific mechanical properties is inherently a multi-objective optimization problem. We present a framework that integrates an active Pareto front learning algorithm (PyePAL) with visualization and explainable AI techniques to optimize processing parameters. PyePAL uses Gaussian process models to predict objective values (hardness and elasticity) from the design variables (spin speed, dilution, and polymer mixture), guiding the adaptive selection of samples toward promising regions of the design space. To enable interpretable insights into the high-dimensional design space, we utilize UMAP (Uniform Manifold Approximation and Projection) for two-dimensional visualization of the Pareto front exploration. Additionally, we incorporate fuzzy linguistic summaries, which translate the learned relationships between process parameters and performance objectives into linguistic statements, thus enhancing the explainability and understanding of the optimization results. Experimental results demonstrate that our method efficiently identifies promising polymer designs, while the visual and linguistic explanations facilitate expert-driven analysis and knowledge discovery.

This work aimed to evaluate both the usefulness and user acceptance of five gradient‐based explainable artificial intelligence (XAI) methods in the use case of a prostate carcinoma clinical decision support system environment. In addition, we aimed to determine whether XAI helps to increase the acceptance of artificial intelligence (AI) and recommend a particular method for this use case. The evaluation was conducted on a tool developed in‐house with different visualization approaches to the AI‐generated Gleason grade and the corresponding XAI explanations on top of the original slide. The study was a heuristic evaluation of five XAI methods. The participants were 15 pathologists from the University Hospital of Augsburg with a wide range of experience in Gleason grading and AI. The evaluation consisted of a user information form, short questionnaires on each XAI method, a ranking of the methods, and a general questionnaire to evaluate the performance and usefulness of the AI. There were significant differences between the ratings of the methods, with Grad‐CAM++ performing best. Both AI decision support and XAI explanations were seen as helpful by the majority of participants. In conclusion, our pilot study suggests that the evaluated XAI methods can indeed improve the usefulness and acceptance of AI. The results obtained are a good indicator, but further studies involving larger sample sizes are warranted to draw more definitive conclusions.

Since the advent of modern computational technologies, libraries and archives have been harnessing the power of computers to produce electronic finding aids for our archival cultural heritage. Today, with the arrival of generative artificial intelligence (specifically, large language models or LLMs), there are new opportunities to post-process these finding aids and enhance access to archival heritage. In this paper, we present a case study of AI-assisted post-processing; we also show how AI can help unlock cultural heritage if combined with interactive data visualization. Our case study focuses on the Grand Ducal Archive of the Medici, partially digitized and electronically cataloged by the Medici Archive Project. We used generative AI to post-process the electronic catalog of the Medici's early modern period correspondence, developing a prototype visual finding aid. Throughout the paper, we present the post-processing steps and introduce the visual finding aid. In conclusion, we critically examine AI's application in the heritage sector and urge GLAM professionals to embrace the technology and open their collections for AI-assisted post-processing.

The integration of Augmented Reality (AR) and Explainable AI (XAI) within Cyber-Physical Systems (CPS) design is transforming the industrial automation landscape. This study explores how combining AR’s immersive visualization with XAI’s decision transparency enhances collaborative design processes in CPS. The AR-XAI platform developed in this research improves team collaboration by offering real-time visual feedback and enabling interactive decision-making. The platform provides interpretable insights into AI-driven decisions, fostering trust among engineers and decision-makers. Key features of the platform include the ability to visualize complex CPS models, facilitating faster iterations, reducing design errors, and improving design accuracy. The integration of XAI ensures transparency in decision-making by offering clear explanations of AI predictions, which is essential for ensuring accountability and building trust in automated systems. Testing with industrial engineers confirmed that the AR-XAI platform significantly improved design outcomes, with a reduction in errors and enhanced team performance compared to traditional design methods. Moreover, the platform enabled faster decision-making and improved collaboration across diverse teams, demonstrating its potential to optimize CPS design workflows. This research provides valuable insights into the role of AR and XAI in advancing Industry 4.0 and paves the way for more advanced integrations of these technologies in industrial settings. Future research should focus on developing scalable solutions for various industrial applications and exploring more sophisticated AR-XAI integrations for emerging fields like smart cities and autonomous manufacturing.

Skin diseases affect millions worldwide, where early and accurate diagnosis is essential for effective treatment. Limited access to dermatologists, particularly in remote areas, often delays proper medical care. This paper presents SkiDi-X, an advanced AI-powered skin disease detection chatbot that integrates ensemble deep learning, explainable AI, and conversational intelligence into a unified diagnostic system. The approach employs a fallback-based ensemble mechanism combining EfficientNetB0, DenseNet121, and ResNet50, ensuring reliable predictions even when model confidence is low. Trained on the HAM10000 and ISIC 2019 datasets, the models achieved accuracies of 89% (EfficientNet), 89% (DenseNet), and 88% (ResNet), demonstrating stable and high-quality performance. To promote transparency, Grad-CAM visualization highlights the key lesion regions influencing decisions, while a Dialogflow-powered chatbot delivers interactive explanations, possible causes, and suggested actions. By merging automated image analysis with human-like interaction, SkiDi-X provides an interpretable, accessible, and user-friendly platform that enhances the reach and impact of teledermatology in early skin disease assessment.

As the nature of cyber threats becomes more sophisticated, Artificial Intelligence (AI)-driven automated security systems have become an essential tool for real-time threat detection and response. Yet, the black-box nature of Deep Learning (DL) models considerably hinders their use in mission-critical cybersecurity operations because they lack interpretability and trust. In this work, we introduce an Explainable AI (XAI)-enabled Cyber Threat Intelligence (CTI) system that increases transparency and decision-making in autonomous security operations. Our solution combines neuro-symbolic reasoning, causal inference, and attention-based visualization to present human-understandable explanations for identified threats. We utilize Transformer (TM)-based anomaly detection with symbolic logic inference to enhance interpretability without compromising high detection accuracy. Experimental tests on benchmark cybersecurity data sets, such as CIC-IDS2017 and UNSW-NB15, illustrate that our XAI-enhanced framework is 97.5% accurate with a 42% decrease in false positives, surpassing baseline black-box models. Our model also enhances the decision-making speed of human analysts by 35%, as corroborated through actual security operation center (SOC) case studies. The findings show that XAI is able to fill the gap between automation and trust from humans, driving adoption in enterprise security and regulatory compliance (e.g., GDPR, NIST AI RMF). This work establishes a new paradigm for transparent, interpretable, and resilient AI-based cybersecurity, making proactive defense measures against zero-day attacks, adversarial attacks, and Advanced Persistent Threats (APTs) possible.

The Research presents a highly intelligent and transparent system of the stylistic variations between more than two authors on a shared writing platform. The proposed Hybrid Transformer Stylometric Explainable Framework (HTSEF), which is implemented in PyTorch, combines deep contextual transformer embeddings and explainable stylometric features to obtain both semantic and linguistic cues of authorial voice. The system uses explainable AI methods like SHAP and attention-based visualization to give clear and human comprehensible logic behind each voice shift detected. The experimental results showed a significant gain in accuracy, precision and recall in comparison to the current models, and they were able to sustain real-time processing efficiency that could be applied to collaborative writing platforms. The results indicate that deep learning plus explainability are practicable in the context of transparent authorship analysis, leading to a higher level of user trust, content honesty, and quality of digital collaboration in general.

This work-in-progress research paper reports findings from introducing an eXplainable AI (XAI) component into AI education within two Finnish middle school classrooms. The educational intervention focused on image classification, a popular technique for introducing learners to supervised machine learning and data-driven thinking. A key moment during these interventions arises when the model does not work as expected: Learners must engage deeply in data-driven analysis to identify what in their data set might explain the unexpected behavior, collect additional data, or curate their dataset, re-train and test their model, and iterate accordingly. To support learners during the “evaluate your model behavior” phase of the ML workflow, we designed and integrated a class activation map (CAM)-based XAI component into our learning tool. This component provides learners with a heatmap-type visualization that highlights which areas of an image contribute the most to the classification result. We conducted classroom interventions with children aged 11 - 16 years, gathered written design diaries and video data, and analyzed them using qualitative content analysis. Our findings indicate that without explicit instruction on how the XAI component works and how to interpret the visualization, the learners found the component more confusing than explanatory. The study suggests future research directions for facilitating XAI-enhanced educational technology in classroom settings.

In today’s rapidly evolving digital landscape, cybersecurity is paramount as organizations face increasingly sophisticated attacks. Artificial intelligence (AI) has become a key tool in detecting and mitigating these threats; however, conventional AI models often operate as “black boxes,” leaving decision processes obscure. Explainable AI (XAI) emerges as a promising solution by illuminating the internal mechanisms of these models, thereby enhancing transparency and trust. This paper explores the integration of XAI into cybersecurity frameworks to improve model transparency and accountability. We examine techniques such as feature importance analysis, surrogate modeling, and visualization methods that reveal how AI systems identify anomalies and flag potential threats. Our analysis demonstrates that making AI decisions interpretable not only supports security experts in understanding and validating automated responses but also aids in regulatory compliance and ethical oversight. Furthermore, enhanced transparency helps in diagnosing biases and vulnerabilities that could be exploited by adversaries, ultimately strengthening the resilience of cybersecurity systems.

The fast integration of AI in key industries like healthcare, banking, and autonomous cars has led to a growing number of people looking for ways to understand and assure accountability for machine learning models. Users have a hard time accepting, trusting, and collaborating with black box models (e.g., deep neural networks and ensemble approaches) since they don't show how they make decisions, even when these models are good at producing predictions. There is a novel way to bridge this gap with explainable AI (XAI) solutions, which reduce complicated systems to more understandable and observable forms. This research delves into many XAI approaches, including tools for data visualization, models that are universally understandable, and model-agnostic methods like SHAP and LIME. The superior knowledge of feature importance, causal links, and decision pathways that XAI possesses allows for more fair algorithmic decision-making, more trustworthy results, and easier debugging. Then, it moves on to discuss topics like consistency, scalability, and the danger of oversimplification. Finding a middle ground between being clear and being honest is crucial. Explainable AI (XAI) is used to transform "black box" models into transparent systems. This lays the groundwork for the ethical deployment of AI in major real-world settings and allows humans and AI to collaborate.

This paper investigates the application of artificial intelligence (AI) techniques to the challenging problem of stock price prediction, with a focus on advanced deep learning architectures and explainable AI (XAI) methods. The study evaluates and compares the predictive performance of Long Short-Term Memory (LSTM) networks, Gated Recurrent Units (GRU), Transformer networks, and eXtreme Gradient Boosting (XGBoost) within a standardized experimental framework. To enhance model interpretability, multiple XAI approaches are employed, including SHapley Additive exPlanations (SHAP), Local Interpretable Model-agnostic Explanations (LIME), and Attention-based visualization. Experimental results demonstrate that the Transformer model achieves superior accuracy in capturing the complex temporal dependencies of financial time series. Moreover, XAI analysis highlights the significant influence of news sentiment and trading volume in driving stock price fluctuations. These findings not only advance the understanding of AI-based financial forecasting but also provide valuable insights into the interpretability of complex predictive models in the context of financial markets.

The proposed organization is an Explainable AI-based Decision Support System (XAI-DSS) that can optimize real-time project management as a combination of SHAP (SHapley Additive exPlanations) as the fundamental explainability procedure and Microsoft Power BI as the visualization and deployment tool. The system uses machine learning to make projections and optimize all critical project variables like the allocation of resources, schedules, and risk issues. SHAP is used to explain and prove the prediction of AI model to build transparency of AI model and ensure trust by related stakeholders. These findings are graphically illustrated in dynamic form using Microsoft Power BI, making the project managers engage with the real-time data and exposition using user-friendly dashboards. This combination is not only healthier in the manner of improving the quality of the decisions made, but also in the sense in that it creates a sense of accountability and user trust on the AI-based recommendations. The identified framework can be viewed as pragmatic and scalable in terms of increasing the importance of having interpretable, real-time decision support within a complex project environment and is the solution that organizations can implement to improve their project management approaches with the view of transparent artificial intelligence.

Medical image diagnosis has witnessed significant advancement through deep learning technologies, particularly Convolutional Neural Networks (CNNs). However, the black-box nature of traditional CNNs poses challenges in clinical adoption where interpretability is crucial. This study presents a novel hybrid CNN architecture integrated with Explainable AI (XAI) techniques to enhance diagnostic accuracy while maintaining interpretability in medical imaging applications. We developed a multi-scale feature extraction framework combining residual connections, attention mechanisms, and gradient-based visualization techniques. The proposed methodology was evaluated on chest X-ray, dermatological, and retinal fundus datasets, achieving superior performance with 94.7% accuracy for pneumonia detection, 92.3% for skin cancer classification, and 91.8% for diabetic retinopathy grading. The integration of LIME, GradCAM, and SHAP visualization techniques provided clinically meaningful explanations, improving physician trust and decision-making confidence by 23.4%. Results demonstrate that hybrid CNN architectures with XAI integration significantly outperform traditional approaches while maintaining computational efficiency and clinical interpretability requirements.

Abstract The major histocompatibility complex (MHC) encodes a range of immune response genes, including the human leukocyte antigens (HLAs) in humans. These molecules bind peptide antigens and present them on the cell surface for T cell recognition. The repertoires of peptides presented by HLA molecules are termed immunopeptidomes. The highly polymorphic nature of the genres that encode the HLA molecules confers allotype-specific differences in the sequences of bound ligands. Allotype-specific ligand preferences are often defined by peptide-binding motifs. Individuals express up to six classical class I HLA allotypes, which likely present peptides displaying different binding motifs. Such complex datasets make the deconvolution of immunopeptidomic data into allotype-specific contributions and further dissection of binding-specificities challenging. Herein, we developed MHCpLogics as an interactive machine learning-based tool for mining peptide-binding sequence motifs and visualization of immunopeptidome data across complex datasets. We showcase the functionalities of MHCpLogics by analyzing both in-house and published mono- and multi-allelic immunopeptidomics data. The visualization modalities of MHCpLogics allow users to inspect clustered sequences down to individual peptide components and to examine broader sequence patterns within multiple immunopeptidome datasets. MHCpLogics can deconvolute large immunopeptidome datasets enabling the interrogation of clusters for the segregation of allotype-specific peptide sequence motifs, identification of sub-peptidome motifs, and the exportation of clustered peptide sequence lists. The tool facilitates rapid inspection of immunopeptidomes as a resource for the immunology and vaccine communities. MHCpLogics is a standalone application available via an executable installation at: https://github.com/PurcellLab/MHCpLogics.

We are interested in the computational study of shock hydrodynamics, i.e. problems involving compressible solids, liquids, and gases that undergo large deformation. These problems are dynamic and nonlinear and can exhibit complex instabilities. Due to advances in high performance computing it is possible to parameterize a hydrodynamic problem and perform a computational study yielding O(TB) of simulation state data. We present an interactive machine learning tool that can be used to compress, browse, and interpolate these large simulation datasets. This tool allows computational scientists and researchers to quickly visualize ‘what-if’ situations, perform sensitivity analyses, and optimize complex hydrodynamic experiments.

As environmental sciences increasingly rely on complex datasets, machine learning (ML) has become crucial for identifying patterns and relationships. However, the integration of ML into workflows can pose challenges due to technical barriers or the time-intensive nature of coding. To address these issues, we developed iMESc, an interactive ML app designed to streamline and simplify ML workflows for environmental data. Developed in R and built on the Shiny platform, iMESc enables the integration of supervised and unsupervised ML methods, along with tools for data preprocessing, visualization, descriptive statistics, and spatial analysis. The Datalist system ensures seamless transitions between analytical workflows, while the “savepoints” feature enhances reproducibility by preserving the analysis state. We demonstrate iMESc’s flexibility with four workflows applied to a case study predicting nematode community structure based on environmental data. The classical statistical approaches, the Redundancy Analysis (RDA) and Piecewise RDA (pwRDA), explained 30.7% and 53%, respectively. The SuperSOM model achieved an R2 of 0.60 for training and 0.291 for testing, identifying spatial patterns across depth zones. Finally, a hybrid model combining an unsupervised SOM and followed by the supervised Random Forest model returned an accuracy of 83.47% for the training and 80.77% for the test, with Bathymetry, Chlorophyll, and Coarse Sand as key predictive variables. IMESc permits the customization of plots and saving the workflows into “savepoints” guarantying reproducibility. iMESc bridges the gap between the complexity of machine learning algorithms and the need for user-friendly interfaces in environmental research. By reducing the technical burden of coding, iMESc allows researchers to focus on scientific inquiry, improving both the efficiency and depth of their analyses.

Access to safe and potable water is critical for public health and socioeconomic development. Traditional methods for predicting water quality often lack accuracy and robustness due to limitations in data availability and model complexity. This study presents a novel approach to predicting water potability by developing an advanced ensemble model and an interactive visualization dashboard. A comprehensive dataset of water quality parameters was collected and preprocessed to ensure data integrity. An ensemble model combining Decision Trees, Random Forest, Gradient Boosting Machines (GBM), XGBoost and Neural Networks was constructed, leveraging the strengths of each algorithm to enhance predictive accuracy. The model achieved an accuracy of 96.7%, precision of 96.7%, recall of 100%, and F1-score of 98.4%, outperforming existing models in the literature. Furthermore, to address contemporary developments in the field of machine learning, a transformer-based model was introduced alongside the existing ensemble and neural network approaches. This integration reflects the heightened prominence of deep learning methods and aligns the study’s methodology with cutting-edge research trends. Cross-validation results further confirmed the model’s robustness, with a mean accuracy of 96.9% and a low standard deviation. The interactive Water Quality Predictive Dashboard developed using the Dash framework, provides real-time predictions and visualizations, allowing stakeholders to input new data and receive immediate feedback on water potability. This tool enhances user accessibility and supports informed decision-making. The study highlights the advantages of the ensemble approach in improving prediction accuracy and reliability, though it also acknowledges the complexity and data dependency challenges. Overall, the developed model and dashboard offer a powerful solution for water quality monitoring with broad applicability for enhancing public health and resource management.

This research predicts flare-ups of autoimmune disease, where the flare-up prediction is performed through the use of multi-omics data. Fast k-Nearest Neighbor’s (k-NN) with SARSA (State-Action-Reward-State-Action) is the machine learning methodology that breaks the model training for the prediction of flares into seven biological features. The first task to be undertaken is concerned with rescaling practitioners' needs in order. Example: all features can be on the same lexical scale. After the Rescaling, the features undergo a ranking as a process of determining the most useful ones for prediction tasks. Then, the features of the fast-kNN training model are used to perform fast training of kNN to allow easy data classification. Lastly, the three steps of SARSA learning, that is, one that shows that data expectation changes and the ordinary expectation of the data can be used to change the prediction model. SARSA learning is used in this situation for the stimulation of SARSA prediction to make the data interactions better for the prediction model explanation. The results suggest that a better combination of Fast-kNN and SARSA can improve the predictions of the target. It may address the need for a better early warning for a flare-up, allowing for better management of the autoimmune disease and treatment.

Text embeddings–mappings of collections of text to points in high-dimensional space–are a common object of analysis. A classic method to visualize these embeddings is to create a nonlinear projection to two dimensions and look for clusters and other structures in the resulting map. Explaining why certain texts cluster together, however, can be difficult. In this paper, we introduce a human-in-the-loop framework for applying machine learning (ML) to this challenge. The framework has two stages: (1) explain, in which we use ML to produce a description of a pattern; and (2) test, in which the user can verify the explanation by entering new text that fits the pattern, and sees where it appears on the map. If the new text is mapped to the original cluster, that is evidence in favor of the ML-generated explanation. We illustrate this process with a visualization application that provides two kinds of explanations: Natural Language Explanations and Contrastive PhraseClouds. Scenarios on exploring academic papers and literary work showcase the benefit of our workflow in discovering related topics and analyzing thematic differences in text.

No abstract available

This research introduces an approach that merges intelligence (AI) machine learning and interactive visualization to enhance network traffic analysis and forecasting. Initially the framework generates network structures that mimic diverse network scenarios and traffic trends. It then educates sophisticated machine learning algorithms to anticipate network behavior and performance based on these simulations. The setup features a user web interface that allows users to monitor real time network dynamics assess performance metrics and engage with the data through interactive tools to apply practical insights, from these models. With the integration of data, predictive analytics and dynamic visualization network administrators now possess a set of tools for, in depth performance monitoring and proactive network administration.

Predicting the future of stock market is essential for financial professionals and investors. This paper proposes a reliable method for better accuracy in stock price prediction by using the combination of interactive visualization tools and machine learning algorithms. The proposed framework majorly depends on Time Series analysis, which is used to identify previous trends and patterns in stock prices. Random Forest are suitable for financial data, which frequently displays nonlinear correlations, because to its capacity to handle high attribute data and it is very efficient in determining the best possible decision boundary. Random Forest will provide accurate predictions

Hyperparameter optimization (HPO), as a central paradigm of AutoML, is crucial for leveraging the full potential of machine learning (ML) models; yet its complexity poses challenges in understanding and debugging the optimization process. We present DeepCAVE, a tool for interactive visualization and analysis, providing insights into HPO. Through an interactive dashboard, researchers, data scientists, and ML engineers can explore various aspects of the HPO process and identify issues, untouched potentials, and new insights about the ML model being tuned. By empowering users with actionable insights, DeepCAVE contributes to the interpretability of HPO and ML on a design level and aims to foster the development of more robust and efficient methodologies in the future.

Wildfires threaten ecosystems, human populations, and global climate balance, for which predicting and visualizing wildland fires with powerful tools is imperative. This paper describes a hybrid physics-machine learning-based wildfire simulation platform that simulates, forecasts, and visualizes real-time fire spread dynamics by using a web-based interface. The suggested system combines physics-based fire spreading algorithms-accounting for wind velocity, terrain height, fuel dampness, and vegetation cover density-with data models trained in PyTorch. Real-time Earth observation data from NASA FIRMS and OpenWeatherMap APIs are used to add situational accuracy and responsiveness. The simulation engine produces spatial and temporal maps of fire spread supported by interactive visualization modules such as 2D heatmaps, 3D terrain visualization, and time-series analytics. They allow users to personalize ignition points, model various weather conditions, and dynamically analyze evacuation paths. The hybrid modeling strategy not only enhances the accuracy of prediction but also allows for scientific insight, emergency planning, and educational demonstration of wildfire behavior. This work emphasizes how unification of computational modeling, artificial intelligence, and interactive visualization can build a robust digital twin for wildfire danger assessment and decision-making support.

In today's dynamic work environments, tracking employee performance and work efficiency in real-time is critical for optimizing productivity and achieving organizational goals. This paper proposes a Machine Learning-Driven System for Real-Time Visualization of Employee Performance and Work Efficiency. The system leverages advanced machine learning algorithms to analyze data from various sources, including task completion rates, time management metrics, and key performance indicators (KPIs). By processing this data in real-time, the system generates interactive dashboards and visual reports that provide managers with actionable insights into employee productivity trends, work patterns, and areas requiring improvement. The proposed system integrates supervised and unsupervised learning techniques to predict performance outcomes and identify anomalies in work behavior. Features such as automated performance tracking, predictive analytics, and customizable visualization tools enhance decision-making processes and foster a data-driven management culture. Additionally, the system supports continuous learning and adaptation, improving its accuracy and relevance over time.This research highlights the system's architecture, data processing pipeline, machine learning models, and visualization components. Experimental results demonstrate the system's effectiveness in enhancing employee performance evaluation, enabling proactive management strategies, and ultimately contributing to organizational efficiency and growth. Keywords: Machine Learning, Real-Time Visualization, Employee Performance, Work Efficiency, Data Analytics, Predictive Analytics, Workforce Management, Productivity Optimization

Machine learning models often learn latent embedding representations that capture the domain semantics of their training data. These embedding representations are valuable for interpreting trained models, building new models, and analyzing new datasets. However, interpreting and using embeddings can be challenging due to their opaqueness, high dimensionality, and the large size of modern datasets. To tackle these challenges, we present WizMap, an interactive visualization tool to help researchers and practitioners easily explore large embeddings. With a novel multi-resolution embedding summarization method and a familiar map-like interaction design, WizMap enables users to navigate and interpret embedding spaces with ease. Leveraging modern web technologies such as WebGL and Web Workers, WizMap scales to millions of embedding points directly in users’ web browsers and computational notebooks without the need for dedicated backend servers. WizMap is open-source and available at the following public demo link: https://poloclub.github.io/wizmap.

Workplace burnout is an increasingly prevalent issue, affecting employee well-being, productivity, and retention rates. Traditional burnout detection methods, such as self-reported surveys, often lack real-time monitoring and objective assessments. This research proposes an interactive data visualization framework integrated with machine learning-driven predictive analytics to facilitate early burnout detection. The study leverages behavioral, physiological, and productivity-related datasets to identify burnout trends, visualize correlations, and predict risk levels. Through interactive heatmaps, time-series analysis, and correlation matrices, the system enables HR professionals and individuals to proactively monitor stress indicators. Additionally, predictive models using Random Forest and LSTM forecast potential burnout risks, allowing for timely interventions. Our results demonstrate that real-time data visualization enhances engagement and improves burnout detection accuracy compared to conventional methods. This research contributes to the growing field of AI-driven mental health monitoring, emphasizing the importance of proactive intervention strategies to ensure workplace well-being.

Recent years have witnessed growing interest in understanding the sensitivity of machine learning to training data characteristics. While researchers have claimed the benefits of activities such as a human-in-the-loop approach of interactive label correction for improving model performance, there have been limited studies to quantitatively probe the relationship between the cost of label correction and the associated benefit in model performance. We employ a simulation-based approach to explore the efficacy of label correction under diverse task conditions, namely different datasets, noise properties, and machine learning algorithms. We measure the impact of label correction on model performance under the best-case scenario assumption: perfect correction (perfect human and visual systems), serving as an upper-bound estimation of the benefits derived from visual interactive label correction. The simulation results reveal a trade-off between the label correction effort expended and model performance improvement. Notably, task conditions play a crucial role in shaping the trade-off. Based on the simulation results, we develop a set of recommendations to help practitioners determine conditions under which interactive label correction is an effective mechanism for improving model performance.

In traffic engineering, cities rely on large detector datasets to manage traffic. Visualizing these big, multi-dimensional datasets poses challenges such as overplotting and dimension reduction, often rendering traditional visualization techniques inadequate. To address this, we added two machine learning (ML) algorithms (Local Outlier Factor algorithm and K-Prototypes clustering) to an interactive time series visualization to improve exploration by both domain experts and non-experts. We used an original detector dataset of a mid-sized German city. Our findings reveal that the ML algorithms greatly enhanced data exploration in these interactive visualizations, particularly for users with limited domain knowledge. This research directly contributes to the design of traffic data analysis tools, offering a foundation for traffic detection hardware and software improvements but also advancing complex dataset visualization in general. It will ultimately lead to more informed decisions, improved traffic management, and has the potential to reduce air pollutants, thus counteracting climate change.

This paper presents a new technique for creating machine-learning exercise materials using spreadsheets. Spreadsheets are often used in teaching machine learning for beginners and non-specialists. However, computation in machine learning models is divided into a learning phase and an inference phase, and conventional spreadsheet-based materials either rely on the software's extensions for the learning phase or can result in huge sheets, which is unsuitable for learners who want to observe it. We realize the learning phase on a spreadsheet that fits within a PC screen, including visualization for better understanding, without using any extensions. Also, this implementation technique makes it possible to execute or pause the learning process interactively. We will show an example of a neural network implementation and discuss the merits and limitations of this technique.

No abstract available

No abstract available

Researchers and end users generally demand more trust and transparency from Machine learning (ML) models due to the complexity of their learned rule spaces. The field of eXplainable Artificial Intelligence (XAI) seeks to rectify this problem by developing methods of explaining ML models and the attributes used in making inferences. In the area of structural health monitoring of bridges, machine learning can offer insight into the relation between a bridge’s conditions and its environment over time. In this paper, we describe three visualization techniques that explain decision tree (DT) ML models that identify which features of a bridge make it more likely to receive repairs. Each of these visualizations enable interpretation, exploration, and clarification of complex DT models. We outline the development of these visualizations, along with their validity by experts in AI and in bridge design and engineering. This work has inherent benefits in the field of XAI as a direction for future research and as a tool for interactive visual explanation of ML models.

Variational Autoencoders are widespread in Machine Learning, but are typically explained with dense math notation or static code examples. This paper presents VAE Explainer, an interactive Variational Autoencoder running in the browser to supplement existing static documentation (e.g., Keras Code Examples). VAE Explainer adds interactions to the VAE summary with interactive model inputs, latent space, and output. VAE Explainer connects the high-level understanding with the implementation: annotated code and a live computational graph. The VAE Explainer interactive visualization is live at https://xnought.github.io/vae-explainer and the code is open source at https://github.com/xnought/vae-explainer.

We describe LineageD—a hybrid web‐based system to predict, visualize, and interactively adjust plant embryo cell lineages. Currently, plant biologists explore the development of an embryo and its hierarchical cell lineage manually, based on a 3D dataset that represents the embryo status at one point in time. This human decision‐making process, however, is time‐consuming, tedious, and error‐prone due to the lack of integrated graphical support for specifying the cell lineage. To fill this gap, we developed a new system to support the biologists in their tasks using an interactive combination of 3D visualization, abstract data visualization, and correctable machine learning to modify the proposed cell lineage. We use existing manually established cell lineages to obtain a neural network model. We then allow biologists to use this model to repeatedly predict assignments of a single cell division stage. After each hierarchy level prediction, we allow them to interactively adjust the machine learning based assignment, which we then integrate into the pool of verified assignments for further predictions. In addition to building the hierarchy this way in a bottom‐up fashion, we also offer users to divide the whole embryo and create the hierarchy tree in a top‐down fashion for a few steps, improving the ML‐based assignments by reducing the potential for wrong predictions. We visualize the continuously updated embryo and its hierarchical development using both 3D spatial and abstract tree representations, together with information about the model's confidence and spatial properties. We conducted case study validations with five expert biologists to explore the utility of our approach and to assess the potential for LineageD to be used in their daily workflow. We found that the visualizations of both 3D representations and abstract representations help with decision making and the hierarchy tree top‐down building approach can reduce assignments errors in real practice.

This study presents a novel approach to real-time wind tunnel data reduction by integrating a JR3 six-axis force-torque sensor with machine learning algorithms. Traditional aerodynamic testing often involves large volumes of raw data from force balances, which require extensive post-processing. This paper proposes a machine learning-based model that accelerates the data reduction pipeline, allowing for near-instantaneous derivation of aerodynamic coefficients from JR3 balance data. The framework includes a synchronized data acquisition module, signal preprocessing, a trained regression model, and an interactive visualization tool. Results show that the proposed system can achieve real-time performance while maintaining high accuracy, significantly reducing the computational and time costs associated with wind tunnel testing.

No abstract available

This article presents a reproducible machine learning methodology for the early prediction of Alzheimer’s disease (AD) using clinical and behavioural data. A comparative analysis of multiple classification algorithms was conducted, with the Gradient Boosting classifier yielding the best performance (accuracy: 93.9 %, F1-score: 91.8 %). To improve interpretability, SHapley Additive exPlanations (SHAP) were integrated into the workflow to quantify feature contributions at both global and individual levels. Key predictive variables such as Mini-Mental State Examination (MMSE), Activities of Daily Living (ADL), cholesterol levels, and functional assessment scores were identified and visualized using SHAP-based insights. A user-friendly, interactive web application was developed using Streamlit, allowing real-time patient data input and transparent model output visualization. This method offers a practical tool for clinicians and researchers to support early diagnosis and personalized risk assessment of AD, thus aiding in timely and informed clinical decision-making. Accurate Prediction: Gradient Boosting model achieved 93.9 % accuracy for early Alzheimer’s detection. Explainability: SHAP values provided interpretable insights into key clinical features. Clinical Tool: A Streamlit-based web app enabled real-time, explainable predictions for users.

For businesses employee retention is a major issue, and forecasting attrition can assist HR departments to put in place proactive measures to lower turnover. Using methods including Random Forest, XGBoost, Decision Tree, Support Vector Classifier (SVC), Logistic Regression, KNearest Neighbors (KNN), and Naive Bayes, this project uses machine learning approaches to study important factors affecting employee departure. The model discovers trends in job satisfaction, workload, career development, and worklife balance trained on the IBM Analytics dataset with 35 characteristics and 1,500 records. Deployed as an interactive Flask based web application, the system includes capabilities for data upload, forecasting, and model performance visualization. This AI driven solution helps HR staff to find early at-risk employees, manage issues efficiently, and enhance staff stability by offering practical insights. By using predictive analytics in HR management, businesses can lower attrition expenses, improve staff engagement, and create a more resilient setting.

Earthquakes pose a significant global threat, making seismic pattern analysis essential for risk assessment and disaster preparedness. This research provides a thorough investigation into earthquake occurrences from 1960 to 2023. By leveraging data mining methodologies, clustering techniques, geospatial visualization, and predictive modeling, the study identifies key seismic trends. Analyzing a dataset comprising more than 28,000 earthquake records, K-Means and DBSCAN clustering methods were employed to classify seismic activity into natural groupings, highlighting strong associations with tectonic boundaries. Geospatial visualization techniques, including interactive heatmaps and global scatter plots, provided insights into earthquake density and high-risk zones. Furthermore, machine learning techniques were implemented to categorize earthquakes based on risk levels, while SARIMA time series forecasting was utilized to predict earthquake magnitude trends extending through 2050. The findings contribute valuable insights for researchers, policymakers, and emergency response teams, enabling improved disaster preparedness strategies and seismic risk mitigation.

Distributed Denial of Service (DDoS) attacks pose a significant threat to network security by overwhelming a target system with a massive volume of traffic, disrupting legitimate user access. This project presents a DDoS Attack Detection System that utilizes machine learning-based traffic analysis to identify potential attack patterns. The system generates synthetic network traffic data, allowing for model training and evaluation. The web-based interface, built using Flask and JavaScript, enables users to upload traffic datasets, visualize network activity through real-time charts, and receive attack alerts. The detection mechanism classifies network traffic as normal or malicious based on key parameters like packet count and unique IP addresses. Additionally, a mitigation feature allows users to block detected malicious IPs, preventing further attacks. This system provides a user-friendly dashboard with dark-themed aesthetics and interactive elements for an intuitive experience. The integration of data visualization, real-time monitoring, and mitigation makes this project a robust solution for enhancing network security against DDoS attacks.

Digital twins (DT) are increasingly used in healthcare to model patients, processes, and physiological systems. While recent solutions leverage visualization, visual analytics, and user interaction, these systems rarely incorporate structured service design methodologies. Bridging service design with visual analytics and visualization can be valuable for the healthcare DT community. This paper aims to introduce the service design discipline to visualization researchers by framing this integration gap and suggesting research directions to enhance the real-world applicability of DT solutions.

Collaboration between health science and visual analytics research is often hindered by different, sometimes incompatible approaches to research design. Health science often follows hypothesis-driven protocols, registered in advance, and focuses on reproducibility and risk mitigation. Visual analytics, in contrast, relies on iterative data exploration, prioritizing insight generation and analytic refinement through user interaction. These differences create challenges in interdisciplinary projects, including misaligned terminology, unrealistic expectations about data readiness, divergent validation norms, or conflicting explainability requirements. To address these persistent tensions, we identify seven research needs and actions: (1) guidelines for broader community adoption, (2) agreement on quality and validation benchmarks, (3) frameworks for aligning research tasks, (4) integrated workflows combining confirmatory and exploratory stages, (5) tools for harmonizing terminology across disciplines, (6) dedicated bridging roles for transdisciplinary work, and (7) cultural adaptation and mutual recognition. We organize these needs in a framework with three areas: culture, standards, and processes. They can constitute a research agenda for developing reliable, reproducible, and clinically relevant data-centric methods.

While generative artificial intelligence (Gen AI) increasingly transforms academic environments, a critical gap exists in understanding and mitigating human biases in AI interactions, such as anchoring and confirmation bias. This position paper advocates for metacognitive AI literacy interventions to help university students critically engage with AI and address biases across the Human-AI interaction workflows. The paper presents the importance of considering (1) metacognitive support with deliberate friction focusing on human bias; (2) bi-directional Human-AI interaction intervention addressing both input formulation and output interpretation; and (3) adaptive scaffolding that responds to diverse user engagement patterns. These frameworks are illustrated through ongoing work on "DeBiasMe," AIED (AI in Education) interventions designed to enhance awareness of cognitive biases while empowering user agency in AI interactions. The paper invites multiple stakeholders to engage in discussions on design and evaluation methods for scaffolding mechanisms, bias visualization, and analysis frameworks. This position contributes to the emerging field of AI-augmented learning by emphasizing the critical role of metacognition in helping students navigate the complex interaction between human, statistical, and systemic biases in AI use while highlighting how cognitive adaptation to AI systems must be explicitly integrated into comprehensive AI literacy frameworks.

Modern Education is not \textit{Modern} without AI. However, AI's complex nature makes understanding and fixing problems challenging. Research worldwide shows that a parent's income greatly influences a child's education. This led us to explore how AI, especially complex models, makes important decisions using Explainable AI tools. Our research uncovered many complexities linked to parental income and offered reasonable explanations for these decisions. However, we also found biases in AI that go against what we want from AI in education: clear transparency and equal access for everyone. These biases can impact families and children's schooling, highlighting the need for better AI solutions that offer fair opportunities to all. This chapter tries to shed light on the complex ways AI operates, especially concerning biases. These are the foundational steps towards better educational policies, which include using AI in ways that are more reliable, accountable, and beneficial for everyone involved.

Human-Centered learning analytics (HCLA) is an approach that emphasizes the human factors in learning analytics and truly meets user needs. User involvement in all stages of the design, analysis, and evaluation of learning analytics is the key to increase value and drive forward the acceptance and adoption of learning analytics. Visual analytics is a multidisciplinary data science research field that follows a human-centered approach and thus has the potential to foster the acceptance of learning analytics. Although various domains have already made use of visual analytics, it has not been considered much with respect to learning analytics. This paper explores the benefits of incorporating visual analytics concepts into the learning analytics process by (a) proposing the Learning Analytics and Visual Analytics (LAVA) model as enhancement of the learning analytics process with human in the loop, (b) applying the LAVA model in the Open Learning Analytics Platform (OpenLAP) to support humancentered indicator design, and (c) evaluating how blending learning analytics and visual analytics can enhance the acceptance and adoption of learning analytics, based on the technology acceptance model (TAM).

The VAST Challenge 2020 Mini-Challenge 1 requires participants to identify the responsible white hat groups behind a fictional Internet outage. To address this task, we have created a visual analytics system named CA2: Cyber Attacks Analytics. This system is designed to efficiently compare and match subgraphs within an extensive graph containing anonymized profiles. Additionally, we showcase an iterative workflow that utilizes our system's capabilities to pinpoint the responsible group.

High stakes decision-making often requires a continuous interplay between evolving evidence and shifting hypotheses, a dynamic that is not well supported by current AI decision support systems. In this paper, we introduce a mixed-initiative framework for AI assisted decision making that is grounded in the data-frame theory of sensemaking and the evaluative AI paradigm. Our approach enables both humans and AI to collaboratively construct, validate, and adapt hypotheses. We demonstrate our framework with an AI-assisted skin cancer diagnosis prototype that leverages a concept bottleneck model to facilitate interpretable interactions and dynamic updates to diagnostic hypotheses.

Three-dimensional clinical gait analysis is essential for selecting optimal treatment interventions for patients with cerebral palsy (CP), but generates a large amount of time series data. For the automated analysis of these data, machine learning approaches yield promising results. However, due to their black-box nature, such approaches are often mistrusted by clinicians. We propose gaitXplorer, a visual analytics approach for the classification of CP-related gait patterns that integrates Grad-CAM, a well-established explainable artificial intelligence algorithm, for explanations of machine learning classifications. Regions of high relevance for classification are highlighted in the interactive visual interface. The approach is evaluated in a case study with two clinical gait experts. They inspected the explanations for a sample of eight patients using the visual interface and expressed which relevance scores they found trustworthy and which they found suspicious. Overall, the clinicians gave positive feedback on the approach as it allowed them a better understanding of which regions in the data were relevant for the classification.

Analysis of microservices' performance is a considerably challenging task due to the multifaceted nature of these systems. Each request to a microservices system might raise several Remote Procedure Calls (RPCs) to services deployed on different servers and/or containers. Existing distributed tracing tools leverage swimlane visualizations as the primary means to support performance analysis of microservices. These visualizations are particularly effective when it is needed to investigate individual end-to-end requests' performance behaviors. Still, they are substantially limited when more complex analyses are required, as when understanding the system-wide performance trends is needed. To overcome this limitation, we introduce vamp, an innovative visual analytics tool that enables, at once, the performance analysis of multiple end-to-end requests of a microservices system. Vamp was built around the idea that having a wide set of interactive visualizations facilitates the analyses of the recurrent characteristics of requests and their relation w.r.t. the end-to-end performance behavior. Through an evaluation of 33 datasets from an established open-source microservices system, we demonstrate how vamp aids in identifying RPC execution time deviations with significant impact on end-to-end performance. Additionally, we show that vamp can support in pinpointing meaningful structural patterns in end-to-end requests and their relationship with microservice performance behaviors.

Information visualization and natural language are intricately linked. However, the majority of research and relevant work in information and data visualization (and human-computer interaction) involve English-speaking populations as both researchers and participants, are published in English, and are presented predominantly at English-speaking venues. Although several solutions can be proposed such as translating English texts in visualization to other languages, there is little research that looks at the intersection of data visualization and different languages, and the implications that current visualization practices have on non-English speaking communities. In this position paper, we argue that linguistically diverse communities abound beyond the English-speaking world and offer a richness of experiences for the visualization research community to engage with. Through a case study of how two non-English languages interplay with data visualization reasoning in Madagascar, we describe how monolingualism in data visualization impacts the experiences of underrepresented populations and emphasize potential harm to these communities. Lastly, we raise several questions towards advocating for more inclusive visualization practices that center the diverse experiences of linguistically underrepresented populations.

Remote sensing vision tasks require extensive labeled data across multiple, interconnected domains. However, current generative data augmentation frameworks are task-isolated, i.e., each vision task requires training an independent generative model, and ignores the modeling of geographical information and spatial constraints. To address these issues, we propose \textbf{TerraGen}, a unified layout-to-image generation framework that enables flexible, spatially controllable synthesis of remote sensing imagery for various high-level vision tasks, e.g., detection, segmentation, and extraction. Specifically, TerraGen introduces a geographic-spatial layout encoder that unifies bounding box and segmentation mask inputs, combined with a multi-scale injection scheme and mask-weighted loss to explicitly encode spatial constraints, from global structures to fine details. Also, we construct the first large-scale multi-task remote sensing layout generation dataset containing 45k images and establish a standardized evaluation protocol for this task. Experimental results show that our TerraGen can achieve the best generation image quality across diverse tasks. Additionally, TerraGen can be used as a universal data-augmentation generator, enhancing downstream task performance significantly and demonstrating robust cross-task generalisation in both full-data and few-shot scenarios.

Counterfactuals -- expressing what might have been true under different circumstances -- have been widely applied in statistics and machine learning to help understand causal relationships. More recently, counterfactuals have begun to emerge as a technique being applied within visualization research. However, it remains unclear to what extent counterfactuals can aid with visual data communication. In this paper, we primarily focus on assessing the quality of users' understanding of data when provided with counterfactual visualizations. We propose a preliminary model of causality comprehension by connecting theories from causal inference and visual data communication. Leveraging this model, we conducted an empirical study to explore how counterfactuals can improve users' understanding of data in static visualizations. Our results indicate that visualizing counterfactuals had a positive impact on participants' interpretations of causal relations within datasets. These results motivate a discussion of how to more effectively incorporate counterfactuals into data visualizations.

We propose a visualization technique that utilizes neural network embeddings and a generative network to reconstruct original data. This method allows for independent manipulation of individual image embeddings through its non-parametric structure, providing more flexibility than traditional autoencoder approaches. We have evaluated the effectiveness of this technique in data visualization and compared it to t-SNE and VAE methods. Furthermore, we have demonstrated the scalability of our method through visualizations on the ImageNet dataset. Our technique has potential applications in human-in-the-loop training, as it allows for independent editing of embedding locations without affecting the optimization process.

We develop NL2INTERFACE to explore the potential of generating usable interactive multi-visualization interfaces from natural language queries. With NL2INTERFACE, users can directly write natural language queries to automatically generate a fully interactive multi-visualization interface without any extra effort of learning a tool or programming language. Further, users can interact with the interfaces to easily transform the data and quickly see the results in the visualizations.

This paper evaluates the visualization literacy of modern Large Language Models (LLMs) and introduces a novel prompting technique called Charts-of-Thought. We tested three state-of-the-art LLMs (Claude-3.7-sonnet, GPT-4.5 preview, and Gemini-2.0-pro) on the Visualization Literacy Assessment Test (VLAT) using standard prompts and our structured approach. The Charts-of-Thought method guides LLMs through a systematic data extraction, verification, and analysis process before answering visualization questions. Our results show Claude-3.7-sonnet achieved a score of 50.17 using this method, far exceeding the human baseline of 28.82. This approach improved performance across all models, with score increases of 21.8% for GPT-4.5, 9.4% for Gemini-2.0, and 13.5% for Claude-3.7 compared to standard prompting. The performance gains were consistent across original and modified VLAT charts, with Claude correctly answering 100% of questions for several chart types that previously challenged LLMs. Our study reveals that modern multimodal LLMs can surpass human performance on visualization literacy tasks when given the proper analytical framework. These findings establish a new benchmark for LLM visualization literacy and demonstrate the importance of structured prompting strategies for complex visual interpretation tasks. Beyond improving LLM visualization literacy, Charts-of-Thought could also enhance the accessibility of visualizations, potentially benefiting individuals with visual impairments or lower visualization literacy.

Visual validation of regression models in scatterplots is a common practice for assessing model quality, yet its efficacy remains unquantified. We conducted two empirical experiments to investigate individuals' ability to visually validate linear regression models (linear trends) and to examine the impact of common visualization designs on validation quality. The first experiment showed that the level of accuracy for visual estimation of slope (i.e., fitting a line to data) is higher than for visual validation of slope (i.e., accepting a shown line). Notably, we found bias toward slopes that are "too steep" in both cases. This lead to novel insights that participants naturally assessed regression with orthogonal distances between the points and the line (i.e., ODR regression) rather than the common vertical distances (OLS regression). In the second experiment, we investigated whether incorporating common designs for regression visualization (error lines, bounding boxes, and confidence intervals) would improve visual validation. Even though error lines reduced validation bias, results failed to show the desired improvements in accuracy for any design. Overall, our findings suggest caution in using visual model validation for linear trends in scatterplots.

Qualitative data can be conveyed with strings of text. Fitting longer text into visualizations requires a) space to place the text inside the visualization; and b) appropriate text to fit the space available. For quantitative visualizations, space is available in area marks; or within visualization layouts where the marks have an implied space (e.g. bar charts). For qualitative visualizations, space is defined in common text layouts such as prose paragraphs. To fit text within these layouts is a function for emerging NLP capabilities such as summarization.

A non-fungible token (NFT) is a data unit stored on the blockchain. Nowadays, more and more investors and collectors (NFT traders), who participate in transactions of NFTs, have an urgent need to assess the performance of NFTs. However, there are two challenges for NFT traders when analyzing the performance of NFT. First, the current rarity models have flaws and are sometimes not convincing. In addition, NFT performance is dependent on multiple factors, such as images (high-dimensional data), history transactions (network), and market evolution (time series). It is difficult to take comprehensive consideration and analyze NFT performance efficiently. To address these challenges, we propose NFTVis, a visual analysis system that facilitates assessing individual NFT performance. A new NFT rarity model is proposed to quantify NFTs with images. Four well-coordinated views are designed to represent the various factors affecting the performance of the NFT. Finally, we evaluate the usefulness and effectiveness of our system using two case studies and user studies.

Misleading visualizations pose a significant challenge to accurate data interpretation. While recent research has explored the use of Large Language Models (LLMs) for detecting such misinformation, practical tools that also support explanation and correction remain limited. We present MisVisFix, an interactive dashboard that leverages both Claude and GPT models to support the full workflow of detecting, explaining, and correcting misleading visualizations. MisVisFix correctly identifies 96% of visualization issues and addresses all 74 known visualization misinformation types, classifying them as major, minor, or potential concerns. It provides detailed explanations, actionable suggestions, and automatically generates corrected charts. An interactive chat interface allows users to ask about specific chart elements or request modifications. The dashboard adapts to newly emerging misinformation strategies through targeted user interactions. User studies with visualization experts and developers of fact-checking tools show that MisVisFix accurately identifies issues and offers useful suggestions for improvement. By transforming LLM-based detection into an accessible, interactive platform, MisVisFix advances visualization literacy and supports more trustworthy data communication.

Whether it is in the form of transcribed conversations, blog posts, or tweets, qualitative data provides a reader with rich insight into both the overarching trends as well as the diversity of human ideas expressed through text. Handling and analyzing large amounts of qualitative data, however, is difficult, often requiring multiple time-intensive perusals in order to identify patterns. This difficulty is multiplied with each additional question or time point present in a data set. A primary challenge then is creating visualizations that support the interpretation of qualitative data by making it easier to identify and explore trends of interest. By combining the affordances of both text and visualizations, WordStream has previously enabled ease of information retrieval and processing of time-series text data, but the data-wrangling necessary to produce a WordStream remains a significant barrier for non-technical users. In response, this paper presents WordStream Maker: an end-to-end platform with a pipeline that utilizes natural language processing (NLP) to help non-technical users process raw text data and generate a customizable visualization without programming practice. Lessons learned from integrating NLP into visualization and scaling to large data sets are discussed, along with use cases to demonstrate the usefulness of the platform.

Categorical data does not have an intrinsic definition of distance or order, and therefore, established visualization techniques for categorical data only allow for a set-based or frequency-based analysis, e.g., through Euler diagrams or Parallel Sets, and do not support a similarity-based analysis. We present a novel dimensionality reduction-based visualization for categorical data, which is based on defining the distance of two data items as the number of varying attributes. Our technique enables users to pre-attentively detect groups of similar data items and observe the properties of the projection, such as attributes strongly influencing the embedding. Our prototype visually encodes data properties in an enhanced scatterplot-like visualization, encoding attributes in the background to show the distribution of categories. In addition, we propose two graph-based measures to quantify the plot's visual quality, which rank attributes according to their contribution to cluster cohesion. To demonstrate the capabilities of our similarity-based approach, we compare it to Euler diagrams and Parallel Sets regarding visual scalability and show its benefits through an expert study with five data scientists analyzing the Titanic and Mushroom datasets with up to 23 attributes and 8124 category combinations. Our results indicate that the Categorical Data Map offers an effective analysis method, especially for large datasets with a high number of category combinations.

Sensemaking on a large collection of documents (corpus) is a challenging task often found in fields such as market research, legal studies, intelligence analysis, political science, computational linguistics, etc. Previous works approach this problem either from a topic- or entity-based perspective, but they lack interpretability and trust due to poor model alignment. In this paper, we present HINTs, a visual analytics approach that combines topic- and entity-based techniques seamlessly and integrates Large Language Models (LLMs) as both a general NLP task solver and an intelligent agent. By leveraging the extraction capability of LLMs in the data preparation stage, we model the corpus as a hypergraph that matches the user's mental model when making sense of the corpus. The constructed hypergraph is hierarchically organized with an agglomerative clustering algorithm by combining semantic and connectivity similarity. The system further integrates an LLM-based intelligent chatbot agent in the interface to facilitate sensemaking. To demonstrate the generalizability and effectiveness of the HINTs system, we present two case studies on different domains and a comparative user study. We report our insights on the behavior patterns and challenges when intelligent agents are used to facilitate sensemaking. We find that while intelligent agents can address many challenges in sensemaking, the visual hints that visualizations provide are necessary to address the new problems brought by intelligent agents. We discuss limitations and future work for combining interactive visualization and LLMs more profoundly to better support corpus analysis.

In this paper, we assess the visualization literacy of two prominent Large Language Models (LLMs): OpenAI's Generative Pretrained Transformers (GPT), the backend of ChatGPT, and Google's Gemini, previously known as Bard, to establish benchmarks for assessing their visualization capabilities. While LLMs have shown promise in generating chart descriptions, captions, and design suggestions, their potential for evaluating visualizations remains under-explored. Collecting data from humans for evaluations has been a bottleneck for visualization research in terms of both time and money, and if LLMs were able to serve, even in some limited role, as evaluators, they could be a significant resource. To investigate the feasibility of using LLMs in the visualization evaluation process, we explore the extent to which LLMs possess visualization literacy -- a crucial factor for their effective utility in the field. We conducted a series of experiments using a modified 53-item Visualization Literacy Assessment Test (VLAT) for GPT-4 and Gemini. Our findings indicate that the LLMs we explored currently fail to achieve the same levels of visualization literacy when compared to data from the general public reported in VLAT, and LLMs heavily relied on their pre-existing knowledge to answer questions instead of utilizing the information provided by the visualization when answering questions.

In the light of recent advances in embodied data visualizations, we aim to shed light on agency in the context of data visualization. To do so, we introduce Data Agency and Data-Agent Interplay as potential terms and research focus. Furthermore, we exemplify the former in the context of human-robot interaction, and identify future challenges and research questions.

The widespread use of Deep Neural Networks (DNNs) has recently resulted in their application to challenging scientific visualization tasks. While advanced DNNs demonstrate impressive generalization abilities, understanding factors like prediction quality, confidence, robustness, and uncertainty is crucial. These insights aid application scientists in making informed decisions. However, DNNs lack inherent mechanisms to measure prediction uncertainty, prompting the creation of distinct frameworks for constructing robust uncertainty-aware models tailored to various visualization tasks. In this work, we develop uncertainty-aware implicit neural representations to model steady-state vector fields effectively. We comprehensively evaluate the efficacy of two principled deep uncertainty estimation techniques: (1) Deep Ensemble and (2) Monte Carlo Dropout, aimed at enabling uncertainty-informed visual analysis of features within steady vector field data. Our detailed exploration using several vector data sets indicate that uncertainty-aware models generate informative visualization results of vector field features. Furthermore, incorporating prediction uncertainty improves the resilience and interpretability of our DNN model, rendering it applicable for the analysis of non-trivial vector field data sets.

We introduce a novel metric for measuring semantic continuity in Explainable AI methods and machine learning models. We posit that for models to be truly interpretable and trustworthy, similar inputs should yield similar explanations, reflecting a consistent semantic understanding. By leveraging XAI techniques, we assess semantic continuity in the task of image recognition. We conduct experiments to observe how incremental changes in input affect the explanations provided by different XAI methods. Through this approach, we aim to evaluate the models' capability to generalize and abstract semantic concepts accurately and to evaluate different XAI methods in correctly capturing the model behaviour. This paper contributes to the broader discourse on AI interpretability by proposing a quantitative measure for semantic continuity for XAI methods, offering insights into the models' and explainers' internal reasoning processes, and promoting more reliable and transparent AI systems.

We share observations and challenges from an ongoing effort to implement Explainable AI (XAI) in a domain-specific workflow for cybersecurity analysts. Specifically, we briefly describe a preliminary case study on the use of XAI for source code classification, where accurate assessment and timeliness are paramount. We find that the outputs of state-of-the-art saliency explanation techniques (e.g., SHAP or LIME) are lost in translation when interpreted by people with little AI expertise, despite these techniques being marketed for non-technical users. Moreover, we find that popular XAI techniques offer fewer insights for real-time human-AI workflows when they are post hoc and too localized in their explanations. Instead, we observe that cyber analysts need higher-level, easy-to-digest explanations that can offer as little disruption as possible to their workflows. We outline unaddressed gaps in practical and effective XAI, then touch on how emerging technologies like Large Language Models (LLMs) could mitigate these existing obstacles.

Why do explainable AI (XAI) explanations in radiology, despite their promise of transparency, still fail to gain human trust? Current XAI approaches provide justification for predictions, however, these do not meet practitioners' needs. These XAI explanations lack intuitive coverage of the evidentiary basis for a given classification, posing a significant barrier to adoption. We posit that XAI explanations that mirror human processes of reasoning and justification with evidence may be more useful and trustworthy than traditional visual explanations like heat maps. Using a radiology case study, we demonstrate how radiology practitioners get other practitioners to see a diagnostic conclusion's validity. Machine-learned classifications lack this evidentiary grounding and consequently fail to elicit trust and adoption by potential users. Insights from this study may generalize to guiding principles for human-centered explanation design based on human reasoning and justification of evidence.

Explainable Artificial Intelligence (XAI) is essential for building advanced machine learning-powered applications, especially in critical domains such as medical diagnostics or autonomous driving. Legal, business, and ethical requirements motivate using effective XAI, but the increasing number of different methods makes it challenging to pick the right ones. Further, as explanations are highly context-dependent, measuring the effectiveness of XAI methods without users can only reveal a limited amount of information, excluding human factors such as the ability to understand it. We propose to evaluate XAI methods via the user's ability to successfully perform a proxy task, designed such that a good performance is an indicator for the explanation to provide helpful information. In other words, we address the helpfulness of XAI for human decision-making. Further, a user study on state-of-the-art methods was conducted, showing differences in their ability to generate trust and skepticism and the ability to judge the rightfulness of an AI decision correctly. Based on the results, we highly recommend using and extending this approach for more objective-based human-centered user studies to measure XAI performance in an end-to-end fashion.

The increasing use of Machine Learning (ML) in sensitive domains such as healthcare, finance, and public policy has raised concerns about the transparency of automated decisions. Explainable AI (XAI) addresses this by clarifying how models generate predictions, yet most methods demand technical expertise, limiting their value for novices. This gap is especially critical in no-code ML platforms, which seek to democratize AI but rarely include explainability. We present a human-centered XAI module in DashAI, an open-source no-code ML platform. The module integrates three complementary techniques, which are Partial Dependence Plots (PDP), Permutation Feature Importance (PFI), and KernelSHAP, into DashAI's workflow for tabular classification. A user study (N = 20; ML novices and experts) evaluated usability and the impact of explanations. Results show: (i) high task success ($\geq80\%$) across all explainability tasks; (ii) novices rated explanations as useful, accurate, and trustworthy on the Explanation Satisfaction Scale (ESS, Cronbach's $α$ = 0.74, a measure of internal consistency), while experts were more critical of sufficiency and completeness; and (iii) explanations improved perceived predictability and confidence on the Trust in Automation scale (TiA, $α$ = 0.60), with novices showing higher trust than experts. These findings highlight a central challenge for XAI in no-code ML, making explanations both accessible to novices and sufficiently detailed for experts.