Research on Establishing Inbound Strategies for Supermarkets based on LSTM and Gaussian Process Regression Modeling

Abstract

 This paper provides an in-depth study on the challenges of vegetable merchandising in fresh produce supermarkets, aiming to provide a comprehensive set of management strategies to optimize supermarket operations. First, the sales volume and sales of six types of vegetables were analyzed by descriptive statistics and the cyclical trend was explored by time series processing; second, good correlations between edibles and aquatic roots and tubers as well as edibles and eggplants were found by plotting correlation matrices and heat maps of Spearman's coefficients. Next, this paper analyzed the relationship between cost-plus pricing and total sales and predicted the total replenishment and pricing of vegetables in the coming week using an LSTM time series forecasting model and evaluated the model performance using root mean square error (RMSE). Finally, a Gaussian regression model was used to predict a small sample of data to develop an optimal replenishment volume and pricing strategy for the superstore, which maximized the superstore's revenue. The results of the study show that the inventory management efficiency of fresh supermarkets can be effectively improved by these methods.