Automatic Pricing and Replenishment Decisions of Vegetable Products Based on Machine Learning

Abstract

The objective of this paper is to address the replenishment challenge caused by the limited shelf life of vegetables and maximize the profit potential of a diverse range of vegetables, based on extensive data analysis encompassing vegetable sales rates and gross profits. In this study, machine learning techniques, including the Spielman correlation coefficient, are employed to derive the demand price regression curve and R^2 value. Subsequently, a random forest regression model is utilized for training and predicting data, while an ARIMA model based on time series data is applied for individual data processing. The expected return under current prices is calculated iteratively until convergence is achieved in terms of optimal expected returns, best prices, and ideal replenishment strategies. The stability of the data is assessed using the ADF test with satisfactory results obtained. This research yields numerous computational outcomes through machine learning methodologies. Among these findings, broccoli emerges as the most profitable vegetable with the highest purchase quantity and sales volume; meanwhile, Yunnan lettuce stands out as having the largest catch quantity. Demand exhibits a predominantly negative correlation with price; furthermore, after incorporating constraints such as "minimum packaging amount," medium-priced dishes yield maximum profitability. These findings hold significant value and practical implications for real-life problem-solving.