Generative heuristics-driven for blockchain-enhanced reputation systems and dynamic optimization in short food supply chains
针对短食品供应链在线平台的声誉系统易被操纵问题，提出结合区块链和生成式AI的深度强化学习超启发式方法，动态优化供应商-客户匹配，实验表明能提升客户满意度、减少合同失败并增加平台收益。
Recently, short food supply chains (SFSCs) through online platform are proposed as alternatives to overcome issues of traditional supply chains. The objective is to deal with fluctuating demand, freshness constraints, and trust issues. Online platforms use reviews of customers to manage the supplier–customer matching. However, such reputation system remains vulnerable to manipulation. To address this issue, blockchain technology (BT) is proposed. This article deals with the dynamic online matching in SFSCs using BT. First, the problem is formulated as a Markov Decision Process (MDP). Then, a rolling horizon (RH) is proposed as a baseline method. A deep reinforcement learning (RL) hyperheuristics (DRLH), where generative artificial intelligence (AI) (GenAI) is used to draw the local search operators, is used to improve the solution quality. Extensive experiments are conducted on real-world SFSCs scenarios. Findings reveal that blockchain-based reputation system enhances the customer satisfaction, reduces contract failures, and increase platform gains.