Genetic Algorithm-Optimized Regression Tree for Automatic Pricing and Replenishment of Vegetable Products

In fresh food supermarkets, vegetable products require efficient replenishment and pricing decisions due to their limited shelf life and quality fluctuations. This paper aims to reveal the sales patterns of vegetables, predict demand and pricing for the upcoming week, and propose replenishment and pricing strategies for individual products. First, the sales data are cleaned and preprocessed, with key factors extracted through feature engineering. Pearson correlation analysis and heatmaps are used to study the relationships between 30 factors, vegetable categories, and total sales. A regression decision tree model (evaluated based on R²) is built to establish the sales-profit relationship regarding sales volume and cost-plus pricing. To maximize revenue, the ARIMA time series model (50 simulated averages) is used to predict the sales and wholesale price for the next 7 days, and genetic algorithms are applied to optimize parameters. Finally, a linear programming model is employed to determine the optimal replenishment quantity and pricing strategy, thus maximizing the supermarket's total revenue.