机器学习农产品供应链管理

本文旨在探讨人工智能技术如何赋能农户，提升其应对市场变化的能力。在当前农业市场信息不对称、小农户决策滞后的背景下，研究基于文献综述与案例分析，系统梳理了AI技术在农业信息感知、生产决策优化、产销对接等环节的应用机制。研究表明，AI通过构建数据驱动的决策支持系统，能够为农户提供精准的市场预测、生产管理与风险预警，显著增强其市场应变能力与经营效益。然而，该过程仍面临数据利用低效、技术成本偏高、农户数字素养不足等挑战。据此，本文提出应通过技术降本、政策支持、人才培育与机制创新等多维度协同，推动AI技术在农业领域的深度应用，为小农户融入现代化农业体系、实现可持续发展提供有效路径。

随着全球化贸易的不断扩展和消费者对生鲜农产品品质要求的提高，高效而稳定的供应链管理成为确保食品安全和满足市场需求的关键。由于生鲜农产品具有易腐性、季节性和地域性等特点，使得供应链系统易面临众多风险。因此，文章深入探讨生鲜农产品供应链中的风险因素，识别出关键的风险点，然后运用SD方法构建动态模型进行模拟，发现在众多可能影响供应链稳定性的因素中，供应风险是最为关键和显著的一个，鉴于此，提出了一套针对性的风险管理策略，为生鲜农产品供应链管理提供了见解和思路。

在乡村振兴背景下，人工智能与农村电商的融合为农业现代化注入新动能。本文分析AI技术在智能生产、精准营销和供应链优化等领域的应用成效，证实其显著促进了农产品上行和消费升级。同时揭示技术成本高、数据安全隐患和农民数字素养不足等现实瓶颈，并从政策扶持、技术防护和人才培育三个维度提出针对性解决方案，为破解农村电商发展困境提供实践路径。

近年来，在政策引导、技术突破和产业应用的多重驱动下，我国人工智能发展成就斐然。吉林省作为我国重要的商品粮基地和特色农产品产区，拥有丰富的玉米、水稻、人参等优质资源。然而，由于传统农产品流通环节繁复、信息不对称、品牌效应弱等问题反而制约了产业效益。这一矛盾凸显了提升本地农产品电商发展水平的紧迫性。本文通过分析吉林农产品现状，结合人工智能优势，为吉林省农产品电子商务发展开辟出新道路，推动当地农产品电商发展。

在生鲜农产品销售渠道中，由于生鲜农产品的特殊性，需要对于市场信息有较强的感知能力以协调供应链，促进生鲜农产品供应链的高效稳定运行。而供应链特别是上游在市场实时信息对接方面存在明显不足问题。为此，本文借助面向全平台公开的评论数据，以提高供应链信息传递的实时性与准确性，利用ERNIE模型进行评论文本的情感分析，获取用户情感特征集，并依据现有数据进行特征工程，保留提取6个特征后，通过PSO优化获取权重组合ELM-GBDT预测模型进行需求趋势预测。研究结果表明，与单一模型相比，组合预测模型显著提高了对生鲜农产品销量需求趋势的预测准确度，同时发现生鲜农产品需求趋势变化存在一个相对稳定的周期，为生鲜农产品供应链采购决策、库存优化决策等提供有价值的参考。

在“数商兴农”战略深入推进的时代背景下，农产品电商迎来发展新契机，但也面临产销协同不畅、供应链抗风险能力弱等难题。本文基于动态能力理论，构建“政策–技术–组织”协同分析框架，探讨农产品电商供应链韧性的提升机制。研究发现：感知阶段，政策通过强制型工具，使CRM (Customer Relationship Management)系统能整合多渠道数据，精准感知市场需求变化；捕捉阶段，政策借助激励型工具，促使IERP (Integrated Enterprise Resource Planning)系统整合生产与库存资源，将市场需求转化为实际运营效益；重构阶段，政策运用能力型工具，保障CRM-IERP系统在突发风险下协同应急，快速调整供应链策略。研究提出“技术–组织双轮驱动”的企业策略与“三力协同”政策路径，强调数据交互、模块化适配及政策工具的动态适配性，为增强供应链韧性提供理论依据与实践参考。

本文聚焦于数字经济下订单农业供应链的协同治理问题，以对契约与关系结合的传统治理范式的分析作为基础，构建电商平台主导的数字治理整合性分析框架，并从数据驱动、算法调度、数字规则与信用以及能力赋能四个方面阐述其核心机制。然后基于平台能力与战略导向，识别出组织协调型、数据驱动型、技术服务型和生态赋能型四种实践中典型的治理模式，还对这四种模式进行比较分析。最后，研究进一步对该治理模式的风险进行评估并提出协同优化路径，以期为电商情境下的订单农业供应链治理提供新思路，并为平台运营与政策设计提供参考。

本研究聚焦电商高速发展背景下鲜活农产品物流需求的精准预测问题，针对传统预测方法时效性弱、精准度不足的痛点，将BP神经网络模型引入物流需求预测领域。研究通过梳理电商交易数据、物流时效数据、鲜活农产品特性数据等核心指标，构建适配鲜活农产品物流需求的BP神经网络预测模型，优化模型的输入层、隐藏层、输出层参数设置，结合电商订单的季节性、区域性特征完成模型训练与验证。针对年度数据样本量不足的问题，本研究尝试提升数据频度至季度级，同时优化鲜活农产品冷链物流需求量计算公式，纳入城市居民消费量、外来调入量及电商上行发货量等关键指标。通过灰色关联分析筛选核心影响因素后，结合经济学原理深入阐释指标与物流需求的关联机制。研究结果表明，优化后的模型可有效捕捉电商驱动下鲜活农产品物流需求的动态变化规律，预测精准度较传统统计方法提升显著。本研究突破了传统预测方法的局限，为电商与鲜活农产品物流产业的协同规划提供了量化参考，也为农产品物流需求预测提供了可落地的技术方案。研究兼具理论价值与实践意义，创新点突出，可支撑物流企业的后续动态安排与相关决策。

全球化推动了跨境电商迅速崛起，改变了传统农产品的国际贸易模式。通过电商平台，小型农户和中小企业能够直接接触国际市场。物联网、大数据、区块链等数字技术的应用，优化了农产品的生产、供应链和物流管理，提升了精准营销能力和产品质量追溯，使得农产品在国际市场中更具竞争力。然而，跨境电商也面临挑战，特别是对于发展中国家的农户，技术门槛高、数字基础设施不足，标准化和认证要求严格。此外，跨境支付系统不统一、汇率波动和各国法律差异，增加了农产品出口的复杂性和成本。展望未来，市场扩展与销售渠道将通过智能化电商平台和全球仓储配送网络得到进一步优化。技术培训、推动全球认证体系标准化、以及区块链技术应用将降低技术和标准化障碍。数字货币和统一支付系统的发展将解决支付与合规问题，推动农产品出口的可持续增长。

在S省生鲜农产品物流需求预测分析领域，采用组合模型支持下的预测方法显得尤为重要。然而，相关数据存在整合不精确、预测精度不高等问题。基于此，通过主成分分析与多元回归模型的建立，探索数据源的高效整理方法，确保了数据质量与可靠性；利用Shepley值法优化组合预测模型的构造，增强了模型的适应性与准确性；以S省生鲜农产品物流为案例，通过数据来源整理、主成分回归模型的建立及需求预测，以及基于Shepley组合模型的预测，展现模型在实际应用中的有效性与准确性；提出具体建议，旨在通过模型优化与技术应用，提升物流需求预测的准确度，为S省生鲜农产品物流的高效管理与发展提供科学依据。

京津冀协同战略等国家战略给物流市场带来发展机遇。人们对食品安全健康的关注与日俱增，同时随着人们消费升级，对农产品质量和新鲜度的广泛关注给物流业带来极大挑战。本文从农产品物流的需求出发，建立农产品物流指标体系、设计基于大数据和因子分析以及多元线性回归融合的预测模型，利用模型对农产品物流需求进行预测和结果分析，发现影响京津冀物流需求的关键因素是宏观经济和货运量，并提出京津冀地区发展应按照地区定位，并大力发展物流和通信基础设施，提供高质量的物流服务等建议。

现今的气候变化和环境问题对农业生产造成了越来越大的影响，而农产品价格的波动对农业生产者和消费者都具有直接的经济影响。通过研究农产品价格预测，可以帮助人们更好地应对环境变化带来的挑战。为提高短期价格预测精度，提出了一个基于变分模态分解(Variational Mode Decomposition, VMD)和长短期记忆网络(Long Short-Term Memory, LSTM)的多通道短期价格预测模型。该模型利用VMD将原始时间序列数据分解为一系列不同特征的模态函数，并对每个模态分量分别使用长短期记忆网络进行特征分析预测，最后对各模态分量下的预测结果进行整合。实例测试结果表明，VMD-LSTM模型的预测准确度高于LSTM，具有不错的实用效果。

本文主要用ARIMA模型研究农产品的销售问题。选取河南省2020-9到2022-12的农产品销售数据为研究样本，首先对所研究的十类农产品进行白噪声检验，分析数据的价值性并将有研究价值的数据筛选出来，之后通过求解参数，分别对苹果、罐头建立ARIMA(1,1,2)模型、ARIMA(1,1,4)模型，并通过标准化残差图、残差的AFC图、残差的Ljung-Box白噪声检验p值来验证模型的合理性，最后利用求解出的ARIMA模型预测未来半年河南省的农产品销售情况。

随着经济的不断发展，人民对安全、健康的农产品需求越来越多，山东省作为农业大省，拥有丰富的农产品资源，肉类、果类、蔬菜类等农产品的产量位居全国前列。本文以山东省农产品的冷链物流为研究对象，从区域经济发展状况、交通运输水平、国外内市场供需情况、产业结构四个方面综合选取了十二个指标，运用主成分分析法和多元回归研究了影响农产品冷链物流需求的若干因素，得到的模型方程拟合程度好，线性关系显著，能够较为精准的预测实际发生值，为山东省农产品冷链物流的发展提供一定参考。

新冠肺炎疫情的突然爆发，不仅影响了我们的生活、学习与工作，更对我国的各个行业造成了巨大的影响。其中，与每个人生活息息相关的农产品价格就因为疫情的发生产生了巨大的波动，不同类型的农产品间价格走势分化。本文从Chebyshev多项式模型出发，通过分析现有数据，预测未来几周猪肉、鸡蛋、蔬菜的价格走势，进而对农产品的产销环节提出一定的建议。

本文研究了棉花期货价格的预测问题，采用了BP神经网络模型和Heston随机波动模型两种模型并进行比较。首先使用BP神经网络模型对样本数据进行训练、验证和测试，并利用该模型对未来五天的棉花期货价格进行预测。其次，使用极大似然估计法估计Heston随机波动模型中的参数，并基于估计后的模型采用滚动时间窗法对样本外五天的棉花期货价格进行预测。最后，引入五种损失函数对两种方法的预测精度进行比较。实证结果表明，Heston随机波动模型的预测效果优于BP神经网络模型。

随着人民对于生活质量的更高追求以及现代物流供应链的不断完善，生鲜农产品供应链网络得到了飞快的发展。但是由于生鲜农产品供应链上下游节点企业易受到外界各因素的干扰，从而存在供应中断风险，影响供应链网络整体的运行。因此，本文首先根据生鲜农产品供应链的特点和业务流程，结合复杂网络理论，构建其网络拓扑结构图。接着采用三角模糊数的方法确定网络拓扑结构图各边的权重，并引入节点重要度贡献矩阵和网络效率来表征各个节点在供应链网络局部和全局中的重要性。基于此，对供应链中的重要节点采取必要的预防措施，来增强供应链的抗风险能力。最后，通过实证分析对生鲜农产品供应链中各个节点的重要性进行计算和排序，验证了该评估方法的有效性和实用性。

超市作为果蔬农产品供应链的末端节点，直接影响着果蔬农产品供应链的整体水平和风险状态。因此本文以超市果蔬农产品供应链风险评估作为研究重点，提出基于贝叶斯网络的超市果蔬农产品供应链风险评估模型，旨在有效控制超市果蔬农产品供应链风险，提升果蔬农产品供应链管理水平。首先通过文献综述法和问卷调查法建立了风险评价指标体系；其次基于解释结构模型构建了超市果蔬农产品供应链风险评价贝叶斯网络模型；最后通过实证研究和多维度的推理分析筛选出了关键风险因素，并提出风险应对策略。通过研究表明，S连锁超市当前果蔬供应链风险发生的概率较小，为6.48%，但果蔬损耗风险概率高达28.7%，需重点关注。本文将解释结构模型与贝叶斯网络理论相结合，应用于超市果蔬农产品供应链风险评估领域，体现了不同风险评价指标间的相互影响关系，为超市果蔬供应链风险评估提供新思路。

随着近年来农业企业面临的复杂金融环境，如何测度农业企业的信用风险是成为了愈来愈重要的问题。本文旨在探索一种基于BP神经网络的农业企业信用风险测度模型，选取了2023年共316家财务数据健全且具有代表性的农业企业，选取了财务结构、偿债能力、盈利能力、运营效率四个一级大类指标的17个二级指标构建各农业企业信用风险评估指标体系。使用BP神经网络和SVM支持向量机进行二元回归，分别对比了XGBoost二元回归模型、分类树模型(DT)、朴素贝叶斯模型(NB)、随机森林(RF)回归模型。结果显示BP神经网络模型对于企业信用风险测度指标拥有更好的回归能力，且在性能和精度上优于其他模型。

No abstract available

Background: Mitigating the negative impacts of climate change and ensuring food security are critical challenges for sustainable development. Potato crops play a key role in global food security, and optimizing their supply chains can improve yields, reduce waste, and stabilize farmer incomes. This study focuses on the potato supply chain in Boyacá, Colombia, aiming to maximize profitability for smallholder farmers through a data-driven approach. Methods: We developed a hybrid framework combining the newsvendor model, Monte Carlo simulation, and machine learning to optimize inventory decisions under uncertain demand and price conditions. Historical data on potato demand and prices were analyzed to fit probability distributions, and simulation scenarios were run for three main potato varieties. Results: The results show that integrating these methods improves inventory decision-making, with the Criolla Colombia variety yielding positive profitability, while the Diacol Capiro and Pastusa Suprema varieties incur losses under current market conditions. The machine learning model enhances predictive accuracy and supports dynamic planning. Conclusions: The findings demonstrate the potential of advanced analytics to reduce waste, support sustainable practices, and inform agricultural policy. The proposed methodology offers a practical decision-support tool for stakeholders and can be adapted to other crops and regions facing similar operational challenges.

Motivated by the increasing interest in machine learning algorithms for data-driven applications in agri-food addressing sustainability issues and by the ongoing discussion on the interpretability and sustainability of such algorithms, we compare congruently the performance of some state-of-the-art techniques and a new version (here proposed for the first time) of Co-Active Neuro-Fuzzy Inference System, equipped with fractional regularization (CANFIS-T for short). To this end, we consider two case studies retrieved from the literature and dealing with two approaches for sustainability development, i.e. ex-ante Life Cycle Assessment and Supply Chain Operations Reference in the agri-food context. Such approaches are set in a data-driven framework and completed by the above-mentioned machine learning techniques. The state-of-the-art techniques from the relevant literature are the ensemble ANFIS, Radial Basis Function Network and Decision Tree. The techniques are compared from the computational, interpretability and energy standpoints. From a formal perspective, we prove what negatively affects the accuracy of ensemble ANFIS. On the basis of the performed experiments, we notice that except for the ensemble ANFIS, all the approaches can be regarded as sustainable, with energy savings over 99%, while only CANFIS-T keeps both good accuracy and interpretability (with up to 4 rules) when the number of input and output variables gets large.

Agribusiness is important to the global economy, but poor supply chain management can lead to losses and reduced customer satisfaction, which is why big data offers an opportunity to improve supply chain management in these sectors through big data-based technology models that focus on capturing, storing and analyzing big data to improve decision making in supply chain management. This big data improves communication and collaboration between the various players in the supply chain, resulting in better coordination and faster response to changing business requirements. Using real-time data and machine learning algorithms that can make better decisions and increase production and delivery efficiency. As a result, the technology model based on big data brings many benefits to supply chain management in the agri-food industry, such as product quality and transportation planning, accurate information, and time-based decisions, improving product quality and customer satisfaction. Concluding that this system can be applied in an optimal way, seeing that it is a system that generates a lot of costs, but in the future it is possible to see the adequate recovery periods.

The organic food industry faces significant challenges in maintaining traceability, ensuring quality, and building consumer trust. This research introduces "Aristech," a block chain-based decentralized framework designed to enhance transparency and security within organic food supply chains. Agri Tech leverages block chain to record immutable data at each stage, while smart contracts automate key operations to ensure product authenticity and minimize fraud. Evaluations demonstrate Agri Tech effectiveness in addressing inefficiencies in conventional supply chains. Future enhancements include integrating IoT devices and machine learning capabilities.

Abstract With the digitalization era, artificial intelligence (AI) is the primary catalyst of agri-food marketing strategies with a significant impact on consumers’ buying behavior and business plans of retailers. This article discusses the role of smart technologies towards optimizing the shopping experience, personalizing offers, and reducing the supply chain. One, collaborative filtering- and content-analysis-based recommendation schemes allow personalized marketing campaigns, demand forecasting, and customer satisfaction improvements. Two, machine learning-augmented dynamic price models provide real-time adaptive pricing for the retailer to achieve maximum profitability along with product offer availability. Three, applications of neural networks and deep learning by recommendation engines improve providing more accurate suggestions to optimize customer experience. AI also enables agri-food product traceability by utilizing blockchain, providing consumers with information and increasing product trust when they buy. Problems concerning ethics, privacy, and potential algorithmic price collusion are also identified by the study. By means of critical literature review, the article puts forward benefits of AI deployment in agri-food retailing and suggests directions for development. The findings emphasize the need for clear rules of the right application of AI, and continued potential for innovation in the domain.

The agri-food supply chain is a complex network enclosing various stakeholders, from farmers to consumers, with multifaceted interactions and dependencies. Traditional supply chain management approaches often need help adapting to dynamic environments and optimising decision-making processes. Deep reinforcement learning is employed by integrating value-based and policy-based models, enhanced by advanced learning techniques, to tackle these challenges. This paper explores applying Deep Reinforcement Learning (DRL) approaches, including Q-learning, Deep Q-Learning (DQL), and the Actor-Critic method, to optimise the efficiency of the agri-food supply chain. The actor-critic model significantly enhances decision-making processes across various supply chain stages by improving efficiency and increasing profit margins. A specific scenario of sugar processing and distribution is incorporated, considering real-world scenarios to validate our model. DRL methods optimise sugar production, storage and distribution, ensuring timely deliveries and enhancing profitability. The models address fluctuating demand and transportation logistics challenges, resulting in a more streamlined and responsive sugar distribution network. The findings reveal that Actor-Critic and DQL methods significantly outperform traditional Q-learning considering product profitability, offering unique advantages in handling complex state-action spaces.

In agri-food supply chains (AFSC), ensuring both agri-food safety for consumers and increased profitability for farmers remains a complex challenge. The intricacies and dynamics of AFSC demand effective traceability and management solutions that existing methods often fall short of providing. To address these challenges, we introduce a novel approach that combines blockchain technology for agri-food traceability with a Deep Belief Network-based secure Block Chain Management named DBCM for profit optimization. The proposed blockchain-based AFSC framework serves as the foundation for product traceability, offering decentralized security for agri-food tracing data within AFSC. This secure and transparent platform ensures that consumers can trust the safety and origin of the products they purchase, while farmers can leverage it to increase their profit margins. For enhancing the security an advanced hybrid cryptographic mechanism named HEDS (Homomorphic Encryption with Digital-Ring Signatures). The objective of the Data-Based Decision Making with Hierarchical Ensemble Decision Support (DBCM with HEDS) method that has been proposed is to arrive at safe data-driven decisions concerning the manufacturing and warehousing of agri-food products. By optimizing these operations, it facilitates higher profits for farmers. Unlike traditional methods, the DBCM-HEDS employs advanced machine learning techniques to adapt to the dynamic nature of AFSC, resulting in improved profit margins. the effectiveness of our blockchain-based framework is rigorously evaluated through extensive simulation experiments conducted in various AFSC environments. Furthermore, the DBCM-HEDS method consistently outperforms traditional methods, consistently delivering higher product profits. This innovative combination of blockchain technology and deep belief network offers a promising solution to the intricate challenges faced by agri-food supply chains, benefiting both consumers and farmers.

This research aims to improve inventory management throughout the agri-food supply chain, through the use of Deep Reinforcement Learning algorithms. Three Deep Reinforcement Learning algorithms, including Deep Q-Networks, Deep Deterministic Policy Gradient, and Proximal Policy Optimization algorithms were implemented and tested in order to evaluate their ability to actively manage inventories and improve the performance of supply chains. The results of the experimental phase offered important information regarding the performance of each Deep Reinforcement Learning algorithm. The Deep Deterministic Policy Gradient algorithm was identified as a viable choice, offering the best results in terms of accuracy to set the optimal inventories for the supply chain and improving the efficiency of the supply chain. The percentage of cost efficiency improved from 92.5% to 95.8% in the case of the DDPG model. The inventory turnover was also improved, surpassing the level of 8.1 units from the original level of 7.3 units which means that the system converts the inventor into sales in less time. The metric for on-time delivery also beneficiated from several improvements, reaching 96.5% form the level of 93.2% return. The quality metrics also registered a significant improvement and reported level of 96.2% after the implementation of the DDPG algorithm and compared to the level of 94.6% prior to the implementation of the algorithm. These results suggest that using such a system will bring beneficial changes to the supply chain and will offer the possibility of implementing a data-driven inventory system based on Deep Reinforcement Learning.

The management of a food supply chain is difficult and complex because of the product's short shelf-life, time-sensitivity, and perishable nature which must be carefully considered to minimize food waste. Temperature-controlled perishable food supply chain provides the highly crucial facilities necessary to maintain the quality and safety of the product. The storage temperature is the most vital factor in maintaining both the quality and shelf-life of a perishable food. Adequate storage temperature control ensures that perishable foods are transported to the end-users in good quality and safe to consume. This paper presents perishable food storage temperature control through mathematical optimal control model where the storage temperature is regarded as the control variable and the deterioration of the perishable food’s quality follows the first-order reaction. The optimal storage temperature for a single perishable food is determined by applying the Pontryagin's maximum principle to solve the optimal control model problem. For multi-temperature commodities supply chain, an unsupervised machine learning (ML) method, called k-means clustering technique is used to determine the temperature clusters for a range of perishables. Based on descriptive analysis, it is observed that the k-means clustering technique is effective in identifying the best suitable storage temperature clusters for quality control of multi-commodity supply chain.

PurposeThe purpose of this paper is to demonstrate the efficacy of machine learning (ML) in managing natural language processing tasks, specifically by developing two ML models to systematically classify a substantial number of food waste interventions.Design/methodology/approachA literature review was undertaken to gather global food waste interventions. Subsequently, two ML models were designed and trained to classify these interventions into predefined supply chain-related groups and intervention types. To demonstrate the use of the models, a meta-analysis was performed to uncover patterns amongst the interventions.FindingsThe performance of the two classification models underscores the capabilities of ML in natural language processing, significantly enhancing the efficiency of text classification. This facilitated the rapid and effective classification of a large dataset consisting of 2,469 food waste interventions into six distinct types and assigning them to seven involved supply chain stakeholder groups. The meta-analysis reveals the most dominant intervention types and the strategies most widely adopted: 672 interventions are related to “Process and Operations Optimisation”, 457 to “Awareness and Behaviour Interventions” and 403 to “Technological and Engineering Solutions”. Prominent stakeholder groups, including “Processing and Manufacturing”, “Retail” “Government and Local Authorities” and “NGOs, Charitable Organisations and Research and Advocacy Groups”, are actively involved in over a thousand interventions each.Originality/valueThis study bridges a notable gap in food waste intervention research, a domain previously characterised by fragmentation and incomprehensive classification of the full range of interventions along the whole food supply chain. To the best of the authors’ knowledge, this is the first study to systematically classify a broad spectrum of food waste interventions while demonstrating ML capabilities. The study provides a clear, systematic framework for interventions to reduce food waste, offering valuable insight for practitioners in the food system, policymakers and consumers. Additionally, it lays the foundation for future in-depth research in the food waste reduction domain.

Agri-food supply chains frequently struggle with high functional costs, slow decision-making and is the backbone of layer-based AI models like LSTM and Random Forest, which limit scalability and effectiveness. To address these challenges, this paper introduces Adaptive Federated Learning-Based Force Chain Optimization (AFL-SCO), an AI-driven frame designed to enhance real-time decision-making while reducing pall reliance. AFL-SCO incorporates Federated Learning, Edge AI and Blockchain, allowing farms and estates to train AI models locally on Spiking Neural Networks (SNNs) and XGBoost, transmitting only essential model updates to the spoils. A tone-optimizing AI model selection medium ensures dynamic switching between Temporal Convolutional Networks (TCN), Motor AI and XGBoost, optimizing prediction accuracy. Blockchain-grounded smart contracts enable secure, automated supplier deals, while AI-powered adaptive logistics proactively acclimate to transportation dislocations, request oscillations and rainfall conditions. Performance evaluation reveals substantial advancements, including an increase in AI delicacy from 78-85 to 88-92, a 50-60 reduction in spoilage costs, decision potentially dropping from 5-10 seconds to just 1-2 seconds and a 3x improvement in scalability. These advancements make AFL-SCO a cost-effective, scalable and high-performance AI-driven result for remodeling agri-food supply chains.

Abstract - agri-food supply chain involves multiple stake- holders and complex processes, often leading to issues such as lack of transparency, data tampering, inefficiency in logistics, and reduced farmer profitability. Traditional centralized systems are vulnerable to manipulation, where false quality or production data can mislead consumers and disrupt market dynamics. To overcome these challenges, this research introduces an integrated framework that combines Blockchain technology with Deep Reinforcement Learning (DRL) to ensure both traceability and intelligent decision-making. The blockchain layer functions as a decentralized ledger that immutably records every transaction, enabling trust, accountability, and data integrity among farmers, distributors, retailers, and consumers. Smart contracts automate key operations such as registration, product transfer, and validation without third-party interference. Complementing this, the DRL-based Supply Chain Management (DR-SCM) model continuously learns from changing market conditions including demand, price trends, and logistics constraints to recommend optimal actions for production scheduling, inventory control, and sales timing. This adaptive intelligence allows farmers to maximize profits, minimize waste, and align supply with consumer demand in real time. Simulation results and performance evaluations indicate that the proposed Blockchain DRL framework significantly enhances transparency, operational efficiency, and profit optimization compared to traditional heuristic and static models. This work demonstrates a sustainable and intelligent approach to modernizing agri-food supply chain management through the synergy of secure distributed systems and advanced learning algorithms. Key Words:  Blockchain, Deep Reinforcement Learning, Agri-Food Supply Chain, Smart Contracts, Traceability, Optimization

Blockchain based agri food supply chains augmented with Deep Reinforcement Learning (DRL) optimize important problems like perishability, seasonality as well as demand variability. The traditional supply chain models, which are mostly static, fail to respond to the dynamic supply chain conditions and hence there is a strong need for the introduction of new ideas. In this work, blockchain is integrated for secure and transparent data management, and data for automation, like inventory management, route planning and price adjustment, is optimized in real time using DRL. The blockchain is a phenomenon of decentralization that improves trust and traceability, and the flexibility of DRL guarantees resource utilization in a way that can adapt to unexpected settings. The proposed system is piloted and a case studied to show its capability of enhancing supply chain resilience, recycle wasted products and scale efficiently. It specifies the transformative outlook of the mixture of blockchain and AI technologies in agricultural and Supply chain administration.

Agricultural products are often subject to seasonal fluctuations in production and demand. Predicting and managing inventory levels in response to these variations can be challenging, leading to either excess inventory or stockouts. Additionally, the coordination among stakeholders at various level of food supply chain is not considered in the existing body of literature. To bridge these research gaps, this study focuses on inventory management of agri-food products under demand and lead time uncertainties. By implementing effective inventory replenishment policy results in maximize the overall profit throughout the supply chain. However, the complexity of the problem increases due to these uncertainties and shelf-life of the product, that makes challenging to implement traditional approaches to generate optimal set of solutions. Thus, the current study propose a novel Deep Reinforcement Learning (DRL) algorithm that combines the benefits of both value- and policy-based DRL approaches for inventory optimization under uncertainties. The proposed algorithm can incentivize collaboration among stakeholders by aligning their interests and objectives through shared optimization goal of maximizing profitability along the agri-food supply chain while considering perishability, and uncertainty simultaneously. By selecting optimal order quantities with continuous action space, the proposed algorithm effectively addresses the inventory optimization challenges. To rigorously evaluate this algorithm, the empirical data from fresh agricultural products supply chain inventory is considered. Experimental results corroborate the improved performance of the proposed inventory replenishment policy under stochastic demand patterns and lead time scenarios. The research findings hold managerial implications for policymakers to manage the inventory of agricultural products more effectively under uncertainty.

In India, annually, 67 million tonnes of food are wasted, costing the country roughly 92000 crores. The motivation behind recycling food waste is the desire to divert waste from landfills. This study aims to manage food wastage in regional Food Supply 7hChains (FSC). In this model, we use modern generation machine learning techniques to track the amount of food waste in FSCs and redirect it for productive usage. The future recommended model is inferred to be effective across a number of areas based on the numerous studies that were conducted.

Food quality and safety are the essential hot issues of social concern. In recent years, there has been a growing demand for real-time food information, and non-destructive testing is gradually replacing traditional manual sensory testing and chemical analysis methods with lagging and destructive effects and has strong potential for application in the food supply chain. With the maturity and development of computer science and spectroscopic techniques, machine learning and hyperspectral imaging (HSI) have been widely demonstrated as efficient detection techniques that can be applied to rapidly evaluate sensory characteristics and quality attributes of food products nondestructively and efficiently. This paper first briefly described the basic concepts of hyperspectral imaging and machine learning, including the imaging process of HSI, the type of algorithms contained in machine learning, and the data processing flow. Secondly, this paper provided an objective and comprehensive overview of the current applications of machine learning and HSI in the food supply chain for sorting, packaging, transportation, storage, and sales, based on the state-of-art literature from 2017 to 2022. Finally, the potential of the technology is further discussed to provide optimized ideas for practical application.

No abstract available

PurposeThe agricultural supply chain is susceptible to disruptive geopolitical events. Therefore, agri-food firms must devise robust resilience strategies to hasten recovery and mitigate global food security effects. Hence, the central aim of this paper is to investigate how supply chains could leverage digital technologies to design resilience strategies to manage uncertainty stemming from the external environment disrupted by a geopolitical event. The context of the study is the African agri-food supply chain during the Russian invasion of Ukraine.Design/methodology/approachThe authors employ strategic contingency and dynamic capabilities theory arguments to explore the scenario and conditions under which African agri-food firms could leverage digital technologies to formulate contingency strategies and devise mitigation countermeasures. Then, the authors used a multi-case-study analysis of 14 African firms of different sizes and tiers within three main agri-food sectors (i.e. livestock farming, food-crop and fisheries-aquaculture) to explore, interpret and present data and their findings.FindingsDownstream firms (wholesalers and retailers) of the African agri-food supply chain are found to extensively use digital seizing and transforming capabilities to formulate worst-case assumptions amid geopolitical disruption, followed by proactive mitigation actions. These capabilities are mainly supported by advanced technologies such as blockchain and additive manufacturing. On the other hand, smaller upstream partners (SMEs, cooperatives and smallholders) are found to leverage less advanced technologies, such as mobile apps and cloud-based data analytics, to develop sensing capabilities necessary to formulate a “wait-and-see” strategy, allowing them to reduce perceptions of heightened supply chain uncertainty and take mainly reactive mitigation strategies. Finally, the authors integrate their findings into a conceptual framework that advances the research agenda on managing supply chain uncertainty in vulnerable areas.Originality/valueThis study is the first that sought to understand the contextual conditions (supply chain characteristics and firm characteristics) under which companies in the African agri-food supply chain could leverage digital technologies to manage uncertainty. The study advances contingency and dynamic capability theories by providing a new way of interacting in one specific context. In practice, this study assists managers in developing suitable strategies to manage uncertainty during geopolitical disruptions.

No abstract available

This study presents a comprehensive analysis of agricultural price volatility forecasting using a hybrid long short-term memory (LSTM)-Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model. Agricultural price volatility poses critical challenges for food security, economic stability, and the livelihoods of millions, particularly in developing countries like India. Accurately forecasting these price fluctuations is vital for effective policymaking and strategic decision-making in agricultural markets. This study investigates the potential of deep learning models, specifically LSTM, and their integration with GARCH for forecasting agricultural commodity price volatility. Using extensive historical price data for 23 commodities across 165 markets in India from February 2010 to June 2024, the proposed hybrid model demonstrates significantly enhanced accuracy and robustness compared to standalone econometric or deep learning models. The results suggest that this hybrid approach effectively addresses price instability, offering improved predictive capabilities. These findings provide valuable implications for policymakers and stakeholders, emphasizing the adoption of advanced machine learning techniques for better market risk management and policy interventions tailored to agricultural price dynamics.

Many factors shape agricultural economic behaviour, but large datasets on crop yields, market price volatilities and climate change — perhaps too much data, in fact — are the most prominent In this regard, the development of models that accurately identify and predict agricultural economic behaviour, is quite topical. In this paper, a pattern recognition method for agricultural economic behaviour based on Long Short-Term Memory Network (LSTM) is presented, its performance is improved through multiple optimisation methods. In prior work, LSTM networks have demonstrated an outstanding performance in processing time series data, making them appropriate to capture dynamic features such as seasonal variations, cyclical patterns, and market trends in agricultural economic activities. Finally, to improve the model performance, this paper combines data preprocessing methods, Dropout regularization technology, and PSO algorithm for hyperparameter tuning. Such optimising features have proven effectively increasing model capacity to find latent-augmented patterns in agriculture econ space and provide more robust and accurate prediction results. Experimental results show that the model proposed in this paper achieves a significant accuracy improvement over traditional methods for agricultural economic forecasting tasks.

ABSTRACT The agricultural products price has been affecting people’s livelihood issues and economic and social security and stability. The cyclical fluctuation of the price of agricultural products often indirectly affects the inner psychological demand of agricultural consumers. If the prices of production and processing industries, which rely on cheap raw materials as basic raw materials, are subject to frequent and abnormal fluctuations, it is likely to cause further widespread concerns about people’s lives, and ultimately lead to a vicious cycle of falling commodity prices. In recent years, as a result of lack of timely authoritative information all the time, the market price of agricultural products appeared a lot of varieties before short-term rise and fall repeatedly phenomenon. This paper attempts to take the quantity of garlic produced and sold by pork in China as the key object of analysis and research, and analyzes the level of market price index and the main factors influencing the price of pork sales and edible garlic demand in China in recent ten years. In information economics, financial market theory, statistical methods and other relevant mathematical model as the main guidance, combined with China’s agricultural prices during the period of policy, select the RBF neural network theory and the analysis methods for technical improvements, in a variety of market factors affecting economic operation under the mixed operation China period in our country agricultural prices technology wave prediction rule, broken Based on the traditional artificial statistics and forecasting model method of agricultural product price, the short-term forecasting model of Chinese agricultural product market price theory based on information technology innovation method was established.

No abstract available

No abstract available

The intricate nature of agricultural price data possesses a formidable challenge in the modeling process, necessitating the careful selection and fine-tuning of methodologies. Deep learning emerges as a potent tool for enhancing the predictive accuracy and understanding the complexities of agricultural prices. The effectiveness of deep learning methodologies in handling the complex patterns of agricultural price datasets was demonstrated using monthly potato (Solanum tuberosum L.) price data collected from the National Horticultural Board across four distinct markets. The study was carried out during 2023 aimed to compare the performance of deep learning models, including Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) with feed forward Artificial Neural Networks (ANN) using the error metrics such as Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE) and Mean Absolute Error (MAE). The GRU model performed best for the Chandigarh (16.26% MAPE) and Delhi (6.09% MAPE) markets where LSTM model showed superior performance in the Dehradun market (17.81% MAPE) and CNN for Shimla market (12.53% MAPE). The error percentage of deep learning models were remarkably low when compared to the machine learning model.

This paper discusses the need for more accurate and efficient technologies for automating melon classification in the agricultural sector. Deep learning, specifically convolutional neural networks and recurrent neural networks, is a promising method for achieving this. The study assesses the accuracy, speed, and scalability of several deep learning models and investigates integrating other technologies like IoT devices and remote sensing to enhance capabilities. Environmental impact and ethical considerations are also addressed. The summary of different deep learning models, along with their file sizes and Input Image, are studied. In conclusion, the smaller models, such as DenseNet121, have lower weights (29.5MB); on the other hand, the larger models, such as InceptionV3 and Xception, have larger file sizes (87.0MB and 83.0MB, respectively). When selecting models based on storage and computational limitations, it is essential to consider these metrics. The size of data and the hardware used, such as a GPU or CPU, can be changed. Distinct characteristics emerge when comparing various models based on weight size, loading time, and accuracy. The conclusion is DenseNet201 model stands out with the highest accuracy at 99.14%, but it also entails the lengthiest loading time (1.19 hours), albeit having a relatively smaller weight size of 74.6MB. In contrast, InceptionV3 demonstrates a respectable accuracy of 96.45% with a shorter loading time (21.53 minutes), despite a larger weight size of 87.00MB. On the other hand, VGG19, despite having the lowest accuracy at 80.20%, has a weight size of 77.21MB and the longest loading time (1.17 hours).

PurposeIn order to solve the problems of difficulty in lending to family farms and the lack of credit products, it is necessary to classify the credit rating of family farms and determine the credit risk level of different family farms, so that agriculture-related financial institutions can implement different credit strategies.Design/methodology/approachA method based on BP neural network model is proposed to measure the weights of credit evaluation indicators of family farms and the linear weighting method and the fuzzy comprehensive evaluation method are used to establish the final credit rating system for family farms.FindingsThe empirical results show that the majority of the 246 family farms in Inner Mongolia have a low CC rating.Originality/valueBy constructing a sound and reasonable credit rating system for family farms, thus providing an objective evaluation of the credit rating of family farms, the credit granting status of agriculture-related financial institutions will be adapted to the reasonable loan demand status of family farm owners, and the quality and level of their credit approval will be continuously enhanced.

Potatoes are essential for meeting the world's dietary requirements due to their high concentration of carbohydrates, vitamins, and minerals. Potato production is plentiful worldwide, but postharvest processing losses have reduced potato quality. Potatoes can be graded manually, which is time-consuming and labor-intensive. In this study, a neural network regression model was developed to predict the volume of potatoes using 2D images, which can be used as a parameter for grading potatoes automatically. The image processing method was used to extract features like the major axis, minor axis, and roundness of 400 potato samples, and each potato's volume was thoroughly measured using the water displacement method, and a dataset was generated. These attributes were used to create the neural network regression model and predict the volume. The results revealed that the mean squared error (MSE) was 83.93, the root mean squared error (RMSE) was 9.16, and R2 was approximately 0.882. The potatoes were categorized into three groups: small, medium, and large, based on their size. The utilized model demonstrated an overall accuracy rate of 91.4%. Specifically, it achieved 90% and 85% accuracy for small-sized and medium-sized potatoes, respectively, whereas the model displayed a decrease in accuracy from 60% to 70% for large-sized potatoes. To summarize, the suggested neural network regression model provided an efficient approach to volume prediction using 2D images to improve the grading process of agricultural products.

Wuyuan green tea is a famous agricultural product in China and a product protected by national geo-graphical indications. The processed green tea also needs to remove impurities, such as stones, tea stalks, etc. However, tea stalks cannot be classified from Wuyuan green tea using photoelectric sorting and 2D image recognition technology since they have similar colors. This paper adopts hyperspectral imaging technology to solve the problem of inaccurate sorting caused by their similar colors. Green tea containing tea stalks was imaged using a visible and near-infrared camera with a wavelength of 400nm-1000nm. What’s more, Principal Component Analysis (PCA) was adopted to reduce the dimension of the col-lected hyperspectral image. And the Convolutional Neural Network (CNN) was used constructively to identify tea stalks in hyperspectral image, the CNN can automatically learn the corresponding features, avoid the complex feature extraction process. The experimental results showed that the recognition accuracy for tea stalks reaches 98.53%. The method has a high recognition rate and can meet the actual production requirements. After field testing, the selection rate is as high as 97.05%.

With demand forecasting we can predict in how much amount consumer or customer need the product. The business will not function well without well-defined demand forecasting. Inadequate future demand forecasting could leads to loss of money and wastage of food. There are several techniques of artificial intelligence and machine learning which could be utilized for prediction of food demand. Authors in the proposed work applied machine learning techniques include Logistic Regression, Decision Tree, Random Forest, Extreme Gradient Boosting, K-Nearest Neighbours, Gaussian Naive Bayes, and Linear Discriminant Analysis for prediction of food demand. For validation of the proposed work authors have calculated accuracy, variance, and error for all the models. The proposed models are also compared and it has been concluded that Linear Discriminant Analysis performed well in terms of accuracy which was calculated to be 71.9%. In future authors will try same models on different datasets for generalisation.

No abstract available

Food industry is one of the most important industries in Thailand. The case-study company is a condiment manufacturer that needs to efficiently manage and plan for their business. One of the most important issues is demand forecasting. The company should precisely forecast their product demands, which will be used for operation planning. This study proposes forecasting models for both short-term and long-term planning for the company’s main condiment products. The proposed models are time series, machine learning and hybrid forecasting models which will be compared with pure time series and machine learning methods. Unlike previous work, this study proposes an innovative hybrid model, i.e., Holt- Winters exponential smoothing and Seasonal Autoregressive Integrated Moving average hybrid with Artificial neural network, which has never been considered previously. The accuracy is measured by mean absolute percentage error (MAPE) where the results are also compared to the method currently used in the company. The results show hybrid forecasting model provides the lowest overall error for both short-term and long-term forecast. The most accurate model from this paper can provide MAPE of 2.07% from short-term forecast and MAPE of 2.20% for long-term forecast (6 months in advance). When comparing with the company’s existing MAPE of 20.05%, the proposed model can increase forecast accuracy effectively.

—Continued global economic instability and uncertainty is causing difficulties in predicting sales. As a result, many sectors and decision-makers are facing new, pressing challenges. In supply chain management, the food industry is a key sector in which sales movement and the demand forecasting for food products are more difficult to predict. Accurate sales forecasting helps to minimize stored and expired items across individual stores and, thus, reduces the potential loss of these expired products. To help food companies adapt to rapid changes and manage their supply chain more effectively, it is a necessary to utilize machine learning (ML) approaches because of ML’s ability to process and evaluate large amounts of data efficiently. This research compares two forecasting models for confectionery products from one of the largest distribution companies in Saudi Arabia in order to improve the company’s ability to predict demand for their products using machine learning algorithms. To achieve this goal, Support Vectors Machine (SVM) and Long Short-Term Memory (LSTM) algorithms were utilized. In addition, the models were evaluated based on their performance in forecasting quarterly time series. Both algorithms provided strong results when measured against the demand forecasting model, but overall the LSTM outperformed the SVM.

: Accurate food demand forecasting is one of the critical aspects of successfully managing restaurants, cafeterias, canteens, and refectories. This paper aims to develop demand forecasting models for a university refectory. Our study focused on the development of Machine Learning-based forecasting models which take into account the calendar effect and meal ingredients to predict the heavy demand for food within a limited timeframe (e.g., lunch) and without pre-booking. We have developed eighteen prediction models gathered under five main techniques. Three Artificial Neural Network models (i.e., Feed Forward, Function Fitting, and Cascade Forward), four Gauss Process Regression models (i.e., Rational Quadratic, Squared Exponential, Matern 5/2, and Exponential), six Support Vector Regression models (i.e., Linear, Quadratic, Cubic, Fine Gaussian, Medium Gaussian, and Coarse Gaussian), three Regression Tree models (i.e., Fine, Medium, and Coarse), two Ensemble Decision Tree (EDT) models (i.e., Boosted and Bagged) and one Linear Regression model were applied. When evaluated in terms of method diversity, prediction performance, and application area, to the best of our knowledge, this study offers a different contribution from previous studies. The EDT Boosted model obtained the best prediction performance (i.e., Mean Squared Error = 0,51, Mean Absolute Erro = 0,50, and R = 0,96).

This paper presents a hybrid method that combines machine learning and time series forecasting model techniques to enhance the precision and reliability of forecasts. After establishing a learning model for high value-added agri-food demand forecasting, we selected the annual demand for a certain high value-added agri-food as the target value. Other economic indicators, namely the National Income Index, population, and residents' consumption level, were chosen as the key influencing factors affecting product demand. We collected actual demand data sequences from Zhejiang Province for the years 2006–2022 as training and test samples. Comparing the forecasting results, it was found that this hybrid algorithm is an appropriate method for forecasting. Compared to other individual models, it demonstrates higher accuracy and lower mean absolute errors in predicting the demand for high value-added agri-food.

Demand forecasting is one of the biggest challenges in supply chain management, especially in the food industry, due to the various variables that affect people's needs, continuous changes in prices, and overall economic factors. Many variables affect product sales and demands, such as promotional offers, seasons, holidays, and cultural events, among many others. Despite the difficulty, supply chain processes can be enhanced by using machine learning, which typically produces better predictions than conventional approaches. This paper proposes a model to improve demand forecasting accuracy for the food industry. More specifically, the model focuses on chocolate products, using data from a local chocolate distributor in Saudi Arabia. The proposed model will take data from sales at normal times of the year and promotional sales, in holiday times for example, and use cutting edge machine learning techniques to accurately forecast supply and demand levels in the chocolate industry.

No abstract available

Accurate demand forecasting has become very significant, especially in the food sector, since many products have a limited lifespan, and improper management can cause the organization to incur enormous waste and loss. This research focuses on the problem of analyzing accurate food demand and its prediction through the application of machine learning techniques. An ensemble technique such as voting regression is employed, leveraging Random Forest Regressor and Gradient Boosting Regressor, which were the top-performing models. By integrating these two techniques using voting regression, we can leverage their complementary strengths to enhance prediction accuracy. The ensemble aggregates the predictions of both models, typically by averaging, to produce a final prediction. This technique can assist in reducing overfitting and capturing complex relationships in the data, resulting in more robust and accurate forecasts of food demand. The outcomes of the R2-score, Root Mean Square Error (RMSE) and Mean Average Error (MAE) are 0.99, 0.01, and 0.00, respectively.

: In this paper, a comprehensive predictive analytics framework for demand forecasting of fast-moving food products is proposed with emphasis on potato chips across different flavours. The study uses a hybrid approach of SARIMAX (Seasonal Autoregressive Integrated Moving Average model extended to include exogenous factors) and Random Forest algorithms to model historical sales data and exogenous factors such as sports events, festivals, holidays, and promotional activities. The framework's architecture incorporates several components: seasonal pattern analysis for the time series, event study to analyse demand shocks and feature construction for improved model fit. These dynamic factors are incorporated to produce detailed demand forecasts and indicate the percentage change in demand trends. This approach to production planning is ahead of time to avoid overproduction, minimize waste and achieve proper inventory management throughout the supply chain. The implementation is a web application developed using Streamlit, with strong user authentication, data handling, and visualization features. The system architecture consists of data ingestion, preprocessing, feature selection, model training, and real-time prediction generation modules. The preprocessing pipeline contains data cleaning algorithms, temporal aggregation and outlier detection that are applied automatically to the data. The SARIMAX model was found to be more accurate in providing point forecasts of demand in real time with an accuracy of 91.47%. This is because it can incorporate both the seasonal components and other variables easily. The framework’s effectiveness was then established through a rigorous cross-validation procedure and the use of standard performance metrics such as MAE and RMSE.

Accurate demand forecasting in the food industry is critical to optimizing supply chain operations, minimizing waste, and maximizing product availability. Traditional forecasting techniques are often incapable of capturing the complexities of consumer behaviour as well as external influences such as business tariffs, weather conditions or shifts in the economy. These hurdles lead to inefficiencies, excess inventory, or stockouts. This study focuses on the application of Machine Learning (ML) techniques to improve the forecasting accuracy of food demand. Using more transformational algorithms obtained at a deeper level, we hope to make predictions more accurately and control the dynamic nature of food demand in a better way. These fluctuations may make traditional methods often inaccurate, based on seasonality, promotions, or shifting consumer preferences. However, machine learning can better handle these changing variables. In this paper, we introduce a comparative study of multiple ML methods, including time series models like ARIMA and Prophet, and regression models such as decision trees and neural networks. These models are used on past sales data coupled with explanatory variables such as the weather forecast and promotional information. Through the application of machine learning, our goal is to offer more accurate, adaptable forecasting solutions, thus enabling improved inventory control, minimized waste, and a streamlined supply chain. Using this method can greatly enhance the precision of demand prediction, which ultimately can be quite useful to the running of foodindustry

Accurate demand forecasting is crucial for reducing food waste, enhancing resilience, and promoting sustainability in food systems. Relying only on historical sales data is inadequate to predict the demand for perishable food products. The remaining shelf-life of food products and their qualities need to be considered as they play a vital role in the forecasting process, leading to accurate and reliable predictions. This paper proposes a four-stage conceptual framework to predict the daily demand for fresh food items, incorporating two critical variables: health class and remaining shelf life. The framework is applied to a real industry case to monitor the freshness quality of bananas, and important variables are collected during run-time. The collected data are then processed and used to train several classification models to classify the health class of bananas into either fresh, ripening, or spoiled. Not very surprisingly as many other application domains, results reveal that the random forest model outperforms other models in predicting the health class of bananas placed in boxes, achieving about 91 % prediction accuracy. However, this paper presents a first research of its kind with the abundance of real data, and specific comprehensiveness and focus on the overdue industrial problem.

Food waste has become a critical global concern, particularly in large-scale food services where inaccurate demand estimation leads to surplus production and loss. This paper presents a machine learning-based predictive model designed to forecast daily food demand with higher precision, thereby reducing waste and optimizing resource utilization. The system integrates multiple regression algorithms including Random Forest Regressor, Gradient Boosting Regressor, and Ridge Regression, trained on a public food demand forecasting dataset. After preprocessing and feature engineering, the Random Forest model achieved the best performance with a Mean Absolute Error (MAE) of 67.43, outperforming Gradient Boosting (MAE = 72.19) and Ridge Regression (MAE = 73.81). The results demonstrate that ensemble methods can effectively capture complex non-linear relationships between features such as center type, meal category, and historical demand. The proposed framework can be adapted by food suppliers, catering platforms, and smart kitchens to enhance production planning, lower operational costs, and contribute to sustainable food management.

Food waste is a serious problem, with approximately one-third of all food produced globally being wasted each year. This issue not only exacerbates food insecurity but also has significant environmental impacts, such as greenhouse gas emissions, land use, water consumption, and loss of biodiversity, as well as economic losses. Economically, food waste represents a substantial loss of resources, including labour, energy, and capital invested in food production, processing, and distribution. This problem is recognized as a global crisis not only due to inefficient use of resources but also because of its impact on food security. With the rapidly growing global population, addressing food waste has become an urgent necessity to ensure sustainable food systems. Machine learning (ML) offers innovative solutions to this challenge by using large datasets and advanced algorithms to predict food demand more accurately, optimize inventory management, and enhance supply chain efficiency. ML has significant potential in reducing food waste because it can better predict future demands based on past data and adjust stock levels accordingly. This is particularly advantageous in managing perishable foods, as they have a higher likelihood of being wasted. Machine learning algorithms can analyze large datasets to more accurately predict food demand, optimize inventory management, and improve supply chain efficiency. These algorithms, categorized into three main approaches supervised learning, unsupervised learning, and reinforcement learning can be used in various ways to reduce food waste.

This study examines how artificial intelligence (AI) is transforming the food supply chain by improving forecasting, demand planning, and operational efficiency. AI technologies like predictive analytics and machine learning enhance accuracy and responsiveness but face challenges such as data quality, system integration, and high costs. By surveying 370 food industry professionals, the research explores the benefits and barriers of AI adoption, providing actionable insights to optimize supply chain processes. The findings aim to support businesses and policymakers in leveraging AI strategically for competitive advantage and supply chain resilience.

Ahstract- The demand forecasting method involves estimating the number of products that customers will buy using previous data. Many industries, including the food sector and retail, employ this prediction exercise. Prediction is essential in restaurants since the majority of fundamental products have a limited shelf life. Demands are influenced by a variety of overt and covert circumstances, including season, area, and others. In this study, machine learning uses multiple data sources, like internal and external data, to forecast the supply of various goods based on demand. Although various machine learning models have already been applied to predicting demand, very limited work has been performed to explain the black-box nature of the model. In this work, an attempt is made to explain the interpretability of the model. Here, we first present the food demand prediction problem as a regression problem and then apply various machine learning models to predict the demand for food. After that, a stacked-based ensemble model is employed to address various concerns coming from the base models, achieving better prediction. Finally, the interpretability is resolved to utilize effective learning techniques like Local Interpretable Model-agnostic Explanations (LIME).

Food wastage is one of the most persistent challenges of the food industry, leading to immense economic and environmental losses. The root cause of such a problem lies in mistaken demand prediction, which means overproduction, waste spoilage, and improper allocation of resources. It needs powerful systems that can accurately predict future demand according to historical trends and contextual information related to it. A predictive system leveraging Long Short-Term Memory networks addresses the challenges of food wastage and inefficient demand forecasting in the food industry. This helps the system to forecast the food quantities and transactions using the historical sales data incorporated with features such as day, month, and item-specific attributes with 89.68% accuracy. The same helps in reducing the wastes and optimization of inventory management and better use of resources, thus yielding major economic and environmental benefits toward sustainable food production.

The importance of forecasting crop yields in agriculture cannot be overstated. The effects of yield forecasting are observed in all the aspects of the supply chain from staffing to supplier demand, food waste, and other business decisions. However, the process is often inaccurate and far from perfect. This paper explores the potential of using expert forecasts to enhance the crop yield predictions of our global-to-local XGBoost machine learning system. Additionally, it investigates the ERA5 climate model’s viability as an alternative data source for crop yield forecasting in the absence of on-farm weather data. We find that, by combining both the expert’s pre-season forecasts and the ERA5 climate model with the machine learning model, we can—in most cases—obtain better forecasts that outperform the growers’ pre-season forecasts and the machine learning-only models. Our expert-informed model attains yield forecasts for 4 weeks ahead with an average RMSE of 0.0855 across all the plots and an RMSE of 0.0872 with the ERA5 climate data included.

No abstract available

This study focuses on production planning in the food manufacturing sector using hierarchical forecasting. The selected case for the focal study represents food products with a common main ingredient used in manufacturing. We employ two scenarios: 1) forecasting aggregated total sales for all products, and 2) forecasting sales for each product separately to calculate the total requirement. We employed three statistical models: autoregressive integrated moving average (ARIMA), exponential smoothing (ETS), and Prophet, and five machine learning models such as linear regression (LR), k-nearest neighbors (KNN), support vector regression (SVR), random forest (RF), and extreme gradient boosting (XGBoost). The key findings highlight that forecasting aggregated total sales for common ingredients outperformed the forecasting for each product individually and got the sum for the overall requirement. Further, we found that island-level forecasts are more accurate than district- and distribution-center-level forecasts. XGBoost performed as the best forecasting model, and MinT outperformed as the best reconciliation approach. Our study contributes to supply chain strategies when products have common ingredients in the manufacturing industry to optimize their resource allocation and production planning. This novel approach contributes to enhancing operational efficiency in food manufacturing.

Accurate demand forecasting has become extremely important, particularly in the food industry, because many products have a short shelf life, and improper inventory management can result in significant waste and loss for the company. Several machine learning and deep learning techniques recently showed substantial improvements when handling time-dependent data. This paper takes the ‘Food Demand Forecasting’ dataset released by Genpact, compares the effect of various factors on demand, extracts the characteristic features with possible influence, and proposes a comparative study of seven regressors to forecast the number of orders. In this study, we used Random Forest Regressor, Gradient Boosting Regressor (GBR), Light Gradient Boosting Machine Regressor (LightGBM), Extreme Gradient Boosting Regressor (XGBoost), Cat Boost Regressor, Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM) in particular. The results demonstrate the potential of deep learning models in forecasting and highlight the superiority of LSTM over other algorithms. The Root Mean Squared Log Error(RMSLE), Root Mean Square Error(RMSE),Mean Average Percentage Error(MAPE), and Mean Average Error(MAE) reach 0.28, 18.83, 6.56%, and 14.18, respectively.

Agriculture is the Backbone of India. India has major portion of cultivational Land. Many Crops have lot of demand in abroad. India is one of the major exporters of Food products other than software. But due to erratic climatic conditions, lack of human Resource and improper decision making in choosing crops, farmers are not able to make profits. Also due to urbanization, Agriculture Profession is losing its importance slowly. This problem has invited many Researchers to identify the issues and apply several Machine Learning approaches like Autoregressive Integrated Moving Average (ARIMA), Decision trees, Long Short-Term Memory (LSTM), K Nearest Neighbor (KNN) etc. for accurate Yield and Crop Price Prediction. This helps the Farmers to select suitable crops and obtain maximum yield and Profit.

Effectively forecasting the demand for transportation resources during grain transportation is a critical task for ensuring food security and optimizing logistics processes. Growing grain production and export volumes puts significant strain on transport infrastructure, requiring the use of modern analytical tools for accurate freight transportation planning. This study provides a comprehensive analysis of the applicability of machine learning methods for forecasting demand for rail transport resources during long-distance grain transportation. The study is based on empirical data on grain transportation on the Russian railway network for the period 2020–2024, with a total sample size of over 58 months. Three groups of algorithms were comparatively tested: recurrent neural networks with long short-term memory, ensemble gradient boosting methods, and classical time series models. The results showed that a hybrid architecture combining LSTM networks for capturing complex temporal dependencies and XGBoost for processing exogenous factors demonstrated the best accuracy, with an average absolute percentage error of 4.73% on the test dataset. The developed model predicts monthly rolling stock demand with an accuracy of up to 94.2%, exceeding the results of traditional econometric approaches by 31.8 percentage points. Practical implementation of the proposed approach can reduce transportation costs by 18-23% by optimizing resource planning and minimizing grain car downtime during periods of seasonal demand fluctuations.

Recent increases in global food demand have made this research and, therefore, the prediction of agricultural commodity prices, almost imperative. The aim of this paper is to build efficient artificial intelligence methods to effectively forecast commodity prices in light of these global events. Using three separate, well-structured models, the commodity prices of eleven major agricultural commodities that have recently caused crises around the world have been predicted. In achieving its objective, this paper proposes a novel forecasting model for agricultural commodity prices using the extreme learning machine technique optimized with the genetic algorithm. In predicting the eleven commodities, the proposed model, the extreme learning machine with the genetic algorithm, outperforms the model formed by the combination of long short-term memory with the genetic algorithm and the autoregressive integrated moving average model. Despite the fluctuations and changes in agricultural commodity prices in 2022, the extreme learning machine with the genetic algorithm model described in this study successfully predicts both qualitative and quantitative behavior in such a large number of commodities and over such a long period of time for the first time. It is expected that these predictions will provide benefits for the effective management, direction and, if necessary, restructuring of agricultural policies by providing food requirements that adapt to the dynamic structure of the countries.

Price fluctuations in horticultural commodities such as pulses, potatoes, and onions present a critical challenge to the agricultural sector, impacting market stability and food security. Traditional forecasting techniques often fail to account for the complexities introduced by dynamic variables, including weather patterns, supply-demand shifts, and international trade policies. Recognizing the potential of Machine Learning in addressing these limitations, this study explores an advanced ML-based framework for accurate commodity price prediction. Through rigorous experimentation and comparative analysis, the proposed model integrates time-series data, supervised learning techniques, and multidimensional datasets to enhance predictive accuracy. The effectiveness of this approach is assessed using historical price data, enabling insights into market trends and volatility. By equipping stakeholders—from farmers to policymakers—with actionable intelligence, this research contributes to the development of more resilient agricultural markets and supports informed decision-making in the pursuit of global food security.

Food wastage is a major issue that affects the economy, society, and environment on a global scale. Over one-third of the food produced worldwide is lost or wasted each year, underscoring the necessity for sustainable food systems. This research proposes an innovative solution that leverages technology to address food wastage reduction. An application is developed with a focus on four key categories: food wastage prediction and analysis, demand prediction and ingredient optimization, forecasting of near-expiry food prices, and accurate recycling method prediction. By incorporating machine learning and predictive modeling, the application provides restaurants and food administrators with comprehensive organizational statistics to effectively mitigate food wastage. The solution offers a data-driven approach to better plan events, reduce waste in the food industry, accurately forecast both demand and ingredient requirements, strategically focus on prices based on shelf life and predict optimal recycling methods. Through this multi-pronged approach, we aim to tackle the complex issue of food wastage and help to establish sustainable food systems.

The unprecedented COVID-19 pandemic has revealed a chaotic vulnerability in the distribution of food service. People around the world were prevented from getting food products, which weakened international relations and placed governments on high alert. Meanwhile, in developed countries, an oversupply of food commodities caused serious food-waste problems, while in developing countries, food scarcity became critical. In addition, about 1 million people worldwide are food-insecure and in need of food assistance. How to effectively allocate food for food service is key to reduce excess surplus and perform food distributions. This chapter considers the complex problem of food distribution networks by integrating artificial intelligence and machine learning techniques in the process of demand-supply forecasting, food distribution allocation, and vehicle routing. The demand data from food banks and the supply-inventory data from food suppliers can only be better predicted through effective analytics and algorithms, and the transportation routes can only be better optimized through effective routing intelligence. In the post-Covid-19 era, the new technologies brought by these artificial-intelligent and machine-learning techniques provide the food industry with greater opportunities for enhancements. Food assistance is essential to society, where a supply surplus cannot solve hunger, but distributions do; however, the transportation of food products from suppliers to food service recipients is complicated, as these services are often provided by multiple food banks and distributors. With the aid of the latest artificial-intelligence and machine-learning algorithms, the food shortage can be addressed and the food waste shortage can be reduced through better demand forecasts and larger allocation efficiencies. To minimize the transportation times and maximize the safety of food recipients, shorter transportation routes are also desirable.

The increasing global demand for food necessitates the adoption of sustainable agricultural practices. Hydroponic farming, while efficient in resource utilization, faces challenges in accurately predicting stock levels and sales due to dynamic, ever-changing factors. This research presents an optimized ensemble framework for forecasting hydroponic stock levels and sales by integrating linear regression (LR), random forest (RF), and XGBoost, further enhanced through an evolutionary algorithm (EA). The proposed framework is evaluated using root mean square error (RMSE) and mean absolute error (MAE), demonstrating significant accuracy improvements over individual models. The ensemble model achieves an RMSE reduction of 43.82% for stock prediction and 55.3% for sales forecasting compared to the best-performing individual model. Additionally, local interpretable model-agnostic explanations (LIME) are employed to offer stakeholders clear insights into decision-making processes, such as identifying "number of harvested crops" and "sales data" as key drivers of prediction outcomes. This framework supports sustainable development goals (SDGs) 9.3, 12.3, and 12.C by promoting resource efficiency, reducing food waste, and improving small-scale farmer market access. Future research will explore real-time data integration for dynamic adaptation and further model enhancements.

Fertilizer consumption is relevant in the agribusiness industry, governments, and research entities worldwide. The prediction of fertilizer consumption is a critical input in the food and organics production chain. Therefore, the increase in its production could be planned adequately without compromising the environment. The fertilizer consumption analysis through time is a big challenge because the available data is scarce, and satisfactory Machine Learning models are difficult to be obtained. Although some initiatives applying machine learning and statistical methods are commonly used to predict fertilizer consumption, a thorough evaluation of data analytics approaches to improve predictions under different step-ahead horizons is needed. We explored ways to optimize the temporal data model construction, considering different approaches through pair combinations between data preprocessing and Machine Learning methods. We evaluated these approaches for the NPK fertilizer real data in the top ten countries that demand it. The obtained results showed that using the proposed analytic tools may be a way to get reliable predictions to plan future demands.

To enhance replenishment and pricing decisions for fresh vegetable products in fresh food supermarkets, this study develops machine learning time series forecasting and operations research models. By examining the historical sales and demand patterns, we investigate the relationships between different categories of vegetables and their sales volumes. We then investigate optimal replenishment and pricing strategies under different demand and supply scenarios. The sales unit price, category name, wholesale price, and loss rate were linked via product coding for data preprocessing, showing no missing values. The sales volumes of six vegetable categories were analyzed using the Kruskal-Wallis non-parametric test, followed by a differential analysis. The results indicated a significant difference between sales volume and category code. Finally, an XGBoost model was established to explore the impact of relevant features on sales volume. This study utilized the magnitude of the R-squared values in SPSS software to assess the goodness of fit and choose suitable models. Subsequently, a replenishment plan for supermarkets was established. Experimental forecasting and demand forecasting were conducted using ARIMA time series analysis to determine the total daily replenishment quantity for the upcoming seven days.

No abstract available

No abstract available

Using a thorough set of environmental data, the “Farm to Folk” initiative forecasts the most appropriate crops for production by leveraging machine learning methods. This data set includes crucial agricultural indicators, including nitrogen levels, phosphorus content, temperature, humidity, pH levels of water, and other pertinent variables. Using advanced data processing techniques, feature engineering methodologies, and iterative model training procedures, the system attempts to accurately predict the optimal crops to be grown under specific environmental conditions. The efficacy of the predictive model is rigorously evaluated against real-world agricultural scenarios and historical crop yield data. By means of this iterative process, Farm-to-People seeks to provide farmers with a consistent decision-support tool that takes several environmental factors into account, thereby empowering them to make wise crop choice. By facilitating precision agriculture through machine learning-driven predictions, this initiative ultimately aims to preagricultural efficiency, optimize resource utilization, and propromoteustainable farming practices.

In this research, aim is to assess the possibility of using the Block chain technology and machine learning in the global food supply chain revolution. Initial applications of PwC study have two main areas of focus: in compliance and certification audit support and monitoring sustainability practices. Random Forest had an accuracy of 92% in compliance and certification, specificity of 89% and 91%, respectively, sensitivity and F1 score to be 90%, which shows it was highly efficient to classify adherence to food safety regulations and certification standards. On the other hand, sustainability monitoring took huge profitability due to DBSCAN clustering, with a Silhouette Score of 0.75, which successfully detected sustainability clusters. The results show that Apriori association rule learning extracted strong relationships between sustainability practices and certifications, due to a high confidence level (0.85) and lift (2.10). t-SNE reduced the data to 2D dimensions which provides a great visualization of sustainability data (Explained Variance Ratio = 0.85, Reconstruction Error = 0.05). The current research lays stress on how Block chain and machine learning can actually transform performance of supply chain efficiency, transparency and sustainability. The encapsulated outputs and the overall process could be applied more generally to broader food supply chains, and complemented with state-of-the-art AI tools for predictive analytics as well as automated decision-making; thus, setting the stage for a new era of logistics and sustainability in logistical practices across the globe.

The United States food supply chain is one of the most complex and interconnected systems in the world, spanning agricultural production, processing, transportation, storage, and retail distribution. While this complexity enables efficiency and scale, it also increases vulnerability to disruptions caused by climate change, labor shortages, geopolitical shocks, transportation failures, cyber threats, and public health crises. Conventional risk management methods, often reactive and siloed, have proven inadequate in predicting and mitigating systemic shocks that have become frequent occurrences in today’s world. This article examines how artificial intelligence (AI) and data analytics can transform risk forecasting in U.S. food supply chains by enabling real-time adaptive, predictive, and prescriptive decision-making. Leveraging machine learning, predictive analytics, and integrated data ecosystems, the paper examines the various stages of the food supply chain, key drivers of disruption, analytical models, data sources, and the benefits of AI-driven risk forecasting. The study concludes that AI-driven risk forecasting offers a powerful pathway toward building a more resilient, transparent, and sustainable U.S. food system.

No abstract available

No abstract available

The agricultural sector plays a crucial role to promote food security, especially in the dimensions of food availability and accessibility for developing economies. However, the agri-food supply chain entails various risks alongside other local and global risks in obtaining food security. This study aimed to examine risks that are faced in the agri-food supply chain to guarantee food supply since there limited studies that assess risks in the agri-food supply chain in relation to food security. Poultry supply chain in Indonesia is chosen as the case study due to its developing market economies and the commodity's importance as a national protein source. A risk matrix diagram was used as the underlying approach to identify, assess, and measure food security-related risks in the agri-food supply chain. Findings of this study are expected to illustrate the risk related to food security in developing countries. Additionally, the research methodology is expected to contribute as a future guide to a systematic risk assessment in the context of the agri-food supply chain and food security.

This study presents an AI-powered approach to enhance quality control and traceability in the agri-food sector, focusing on the automated detection and classification of two Tunisian date varieties: Deglet Nour and "Bsir". The main objective is to develop a smart system that can quantitatively and qualitatively determine the proportion of any contamination of one variety by the other within a batch. To achieve this, state-of-the-art object detection YOLO models, v8 and v12, have been employed, trained on a custom annotated dataset which includes a wide range of real-world images, capturing the variability in the studied date fruit size, shape, and presentation. Both YOLO models were fine-tuned over 50 epochs using transfer learning techniques, allowing them to adapt effectively to the specific classification task. Training step consisted of a thorough analysis of bounding box distributions and samples clustering, taking into account natural variations in date morphology based on their 2D images. Evaluation showed that both models achieved high detection accuracy, with YOLOv12 outperforming slightly in precision and speed, making it well-suited for real-time applications. By estimating the relative variety proportions within a mixed batch, the developed smart system supports the intelligent decision-making across the supply chain. This work lays the groundwork for embedding deep learning models into portable smart optical devices that can assess date mixtures on-site, from farms to packaging centers. Future developments will focus on expanding detection to additional date varieties and integrating the system into commercial post-harvest processes.

ABSTRACT The COVID-19 pandemic has underscored the critical importance of resilience in the agri-food supply chain (AFSC) to ensure a reliable, uninterrupted, and disruption-free food supply for the ever-increasing global population. This study aims to comprehensively rank the multitude of risks faced by the AFSC, which have predominantly been examined in isolation. It develops a strategic framework for mitigating these risks by implementing appropriate resilient strategies. The study employs novel hybrid multi-criteria decision-making techniques, including the fuzzy analytical hierarchy process (F-AHP), fuzzy technique for order preference by similarity to ideal solution (F-TOPSIS), and fuzzy quality function deployment (F-QFD), to prioritise risks, assign weights, and rank various resilient strategies. The study contributes to the existing literature by proposing a systematic and robust strategic framework to facilitate decision-making processes in AFSC resilience management. It advances theoretical understanding and provides practical insights and recommendations for enhancing resilience in AFSC management.

No abstract available

Globalization and multinational commerce have increased the dynamism and complexity of supply networks, thereby increasing their susceptibility to disruptions along interconnected supply chains. This study aims to tackle the significant concern of supplier selection disruptions in the Thai agri-food industry as a response to the aforementioned challenges. A novel supplier evaluation system, PROMETHEE II, is suggested; it combines the Fuzzy Analytical Hierarchy Process (FAHP) with inferential statistical techniques. This investigation commences with the identification of critical indicators of risk in the sustainable supply chain via three phases of analysis and 315 surveys of management teams. Exploratory factor analysis (EFA) is utilized to ascertain six supply risk criteria and twenty-three sub-criteria. Following this, the parameters are prioritized by FAHP, whereas four prospective suppliers for an agricultural firm are assessed by PROMETHEE II. By integrating optimization techniques into sensitivity analysis, this hybrid approach improves supplier selection criteria by identifying dependable solutions that are customized to risk scenarios and business objectives. The iterative strategy enhances the resilience of the agri-food supply chain by enabling well-informed decision-making amidst evolving market dynamics and chain risks. In addition, this research helps agricultural and other sectors by providing a systematic approach to selecting low-risk suppliers and delineating critical supply chain risk factors. By bridging complexity and facilitating informed decision-making in supplier selection processes, the results of this study fill a significant void in the academic literature concerning sustainable supply chain risk management.

The agri-food sector is subject to various sources of uncertainty and risk that can have a negative impact on its supply chain performance if not properly managed. In order to determine what actions the supply chain (SC) should take to protect itself against risks, it is necessary to analyze whether the supply chain is robust to them. This paper proposes a tool based on a system dynamics model to determine the robustness of an already designed five-stage fresh agri-food supply chain (AFSC) and its planting planning to disruptions in demand, supply, transport, and the operability of its nodes. The model is validated using the known behavior replication test and the extreme conditions test. In order to guide decision-makers in the different uses of the above system dynamic model, a methodology for the improvement of the AFSC robustness is presented and applied to a case study. As a result, the SC robustness to the defined disruptions is provided. For critical disruptions, protective actions are defined. Finally, the model is re-run to evaluate the impact of these proactive strategies on the AFSC in order to finally select the most beneficial for improving its robustness.

Agri-food supply chains (AFSCs) are becoming more complex in structure, and thus more susceptible to different vulnerabilities and risks. Therefore, to enhance performance, we need to manage the risks in AFSCs effectively and efficiently. This study analyses various AFSC risks using a multi-method approach, including thematic analysis, total interpretive structural modelling (TISM) and fuzzy cross-impact matrix multiplication applied to classification (MICMAC) analysis. Based on the empirical data collected from experienced AFSC practitioners and following thematic analysis, eight categories of risk and 16 risk factors were identified as important. Furthermore, the interrelationships among the identified risks were built using TISM. Finally, the identified risks were classified into various categories according to their dependence and driving power using fuzzy MICMAC analysis. The research results indicate that the weather-related and political risks have the highest driving power and are located at the lowest level in the TISM hierarchy. These risks have a high tendency to disturb the whole flow of AFSC and so should be managed effectively. This study advances existing literature on identifying risk factors, defining interrelations between different AFSC risks, and determining the key risks. The risk analysis results can help AFSC practitioners in AFSC to identify, categorise and analyse the risks.

This paper focuses on the challenges in food safety governance in megacities, taking Shanghai as the research object. Aiming at the pain points in food sampling inspections, it proposes a risk prediction and regulatory optimization scheme combining text mining and machine learning. First, the paper uses the LDA method to conduct in-depth mining on over 78,000 pieces of food sampling data across 34 categories in Shanghai, so as to identify core risk themes. Second, it applies SMOTE oversampling to the sampling data with an extremely low unqualified rate (0.5%). Finally, a machine learning prediction model for food safety risks is constructed, and predictions are made based on this model. The research findings are as follows: ① Food risks in Shanghai show significant characteristics in terms of time, category, and pollution causes. ② Supply chain links, regulatory intensity, and consumption scenarios are among the core influencing factors. ③ The traditional “full coverage” model is inefficient, and resources need to be tilted toward high-risk categories. ④ Public attention (e.g., the “You Order, We Inspect” initiative) can drive regulatory responses to improve the qualified rate. Based on these findings, this paper suggests that relevant authorities should ① classify three levels of risks for categories, increase inspection frequency for high-risk products in summer, adjust sampling intensity for different business entities, and establish a dynamic hierarchical regulatory mechanism; ② tackle source governance, reduce environmental pollution, upgrade process supervision, and strengthen whole-chain risk prevention and control; and ③ promote public participation, strengthen the enterprise responsibility system, and deepen the social co-governance pattern. This study effectively addresses the risk early warning problems in food safety supervision of megacities, providing a scientific basis and practical path for optimizing the allocation of regulatory resources and improving governance efficiency.

PurposeSupply chain risk management can effectively reduce the loss of retailers. In this regard, retailers need to consider the competition risks of competitors in addition to the disruption risks. This paper designs a resilient retail supply chain network for perishable foods under the dynamic competition to maximize retailer's profits.Design/methodology/approachA two-stage mixed-integer non-linear model is presented for designing the supply chain network. In the first stage, an equilibrium model that considers the characteristics of perishable foods is developed. In the second stage, a mixed integer non-linear programming model is presented to deal with the strategic decisions. Finally, an efficient memetic algorithm is designed to deal with large-scale problems.FindingsThe optimal the selection of suppliers, distribution centers and the order allocation are found among the supply chain entities. Considering the perishability of agri-food products, the equilibrium retail price and selling quantity are determined. Through a numerical example, the optimal inventory period under different maximum shelf life and the impact of three resilient strategies on retailer's profit, selling price and selling quantity are analyzed.Research limitations/implicationsAs for future research, the research can be extended in a number of directions. First, this paper studies the retail supply chain network design problem under competition among retailers. It can be an interesting direction to consider retailers competing with suppliers. Second, the authors can try to linearize the non-linear model and solve the large-scale integer programming problem by exact algorithm. Finally, the freshness of perishable foods gradually declines linearly to zero as the maximum shelf life approaches, and it would be a meaningful attempt to consider the freshness of perishable foods declines exponentially.Originality/valueThis paper innovatively designs the resilient supply chain network for perishable foods under dynamic competition. The retailer's dynamic competition and resilient strategies are considered simultaneously when designing supply chain network for perishable foods. In addition, this paper gives insights into how to obtain the optimal inventory period and compare the retailer's resilient strategies.