A Systematic Review of Artificial Intelligence and Predictive Analytics for Dynamic Pricing and Inventory Optimization in Perishable Supply Chains

Perishable supply chains face persistent operational challenges, including demand uncertainty, product deterioration, inventory imbalance, waste generation, and limited real-time visibility. This systematic review examines recent research on artificial intelligence and predictive analytics for demand forecasting, quality and condition assessment, inventory optimization, dynamic pricing, and supply chain resilience in perishable-product environments. The reviewed literature encompasses deep reinforcement learning methods such as DQN, PPO, A۲C, DDPG, and P-DQN; temporal and hybrid forecasting models such as LSTM and ARIMA-LSTM; ensemble learning approaches including Random Forest, Gradient Boosting, and CatBoost; and enabling technologies such as IoT-based sensing, blockchain-based traceability, and computer vision models including VGG-۱۶. The evidence indicates that AI-driven approaches can improve forecasting accuracy, support adaptive pricing and inventory decisions, enhance condition monitoring, and contribute to waste reduction and operational efficiency. Nevertheless, the literature remains fragmented, with limited real-world validation, weak scalability, insufficient interpretability, and inadequate integration across forecasting, sensing, pricing, and decision-making layers. This review identifies these limitations and proposes a future research direction that combines temporal modeling, offline reinforcement learning, federated learning, IoT sensing, and explainable AI to enable scalable and privacy-aware supply chain optimization across perishable sectors.

کلیدواژه ها:

Artificial Intelligence ، Predictive Analytics ، Perishable Supply Chains ، Demand Forecasting ، Dynamic Pricing ، Inventory Optimization ، Reinforcement Learning ، Waste Reduction