Research on Automatic Vegetable Pricing Strategies Based on Multiple Machine Learning Algorithms

Abstract

This paper explores methods for optimizing product pricing and replenishment strategies in fresh supermarkets through advanced data analysis techniques. The initial data processing involves merging and cleaning, with a focus on statistical and visual analysis of sales distribution across products and categories. Descriptive statistics are derived using line charts, bar charts, and pie charts based on yearly, quarterly, and monthly data. This article analyzes relationships within the data employing the Kruskal-Wallis test, Spearman correlation coefficient, and Pearson correlation coefficient. Spectral clustering is utilized to group products, revealing patterns in their interrelationships. For pricing strategies, this paper implements a cost-plus pricing model to calculate markup rates, considering sales and fixed costs. This article apply regression algorithms, specifically XGBoost and Random Forest, to predict unit price, purchase price, and sales volume, and assess their performance for optimization. Time series forecasting techniques predict data for the upcoming week, complemented by an optimization function that visualizes this iterative process. Further data processing for optimizing purchase quantities employs regression algorithms, time series models, and genetic optimization algorithms to refine predictions, ultimately identifying strategies to maximize profit.