Optimizing Replenishment and Pricing Strategies for Vegetable Categories in Supermarkets: A Gradient Boosting and SARIMA Approach

This paper presents a comprehensive analysis of replenishment and pricing strategies for vegetable categories in supermarkets. The study begins by employing gradient boosting tree models to investigate the relationship between total sales volume and cost-based pricing for six vegetable categories: flowers and leaves, cruciferous vegetables, aquatic roots and stems, solanaceous vegetables, chili peppers, and edible fungi. The distribution variations of these categories are examined to understand their market dynamics. Subsequently, the study utilizes K-Fold cross-validation to determine the optimal parameters for the gradient boosting model, evaluating the fitting performance using Mean Squared Error (MSE). In the second part, the research aims to forecast the weekly replenishment quantities and pricing strategies for each vegetable category while maximizing benefits. To achieve this, the study incorporates seasonal time series modeling using SARIMA (Seasonal ARIMA). Following a stationarity test (ADF test), the model solves for the sales volumes and cost-based forecasts for the six vegetable categories from July 1, 2023, to July 11, 2023. A linear regression model is established to analyze the relationship between cost-based pricing and sales volume. The predicted sales volume is treated as the supply quantity, and an optimal pricing strategy is derived to maximize profits by identifying the combination of cost-based pricing and sales volume.