Research on supply chain management and ordering strategies for perishable goods in farmers' markets

Given the uncertainty in market demand and the complexity of supply chain management for perishable vegetable products, traditional demand forecasting and pricing strategies face substantial challenges in practical applications. To address these issues, this study proposes a KAN-LSTM-based framework for demand forecasting and ordering optimization. By integrating the nonlinear representation capability of KAN with the long-term dependency modeling ability of LSTM, the proposed framework provides an effective approach for forecasting perishable vegetable sales and procurement prices. In addition, a multivariate stepwise regression model is employed to estimate price elasticity and support pricing decisions. The ordering strategy is then optimized using the McCormick envelope method combined with the SLSQP algorithm. Experimental results show that the KAN-LSTM model achieves high forecasting accuracy and stable performance in both sales and price prediction tasks. Compared with xLSTM, TCN, and Transformer models, KAN-LSTM demonstrates favorable predictive performance for the dataset considered in this study. The optimization results further indicate that the proposed framework is effective in improving profit and cost control, highlighting its practical value for perishable goods supply chain management in farmers’ markets.