Automatic Pricing and Replenishment Decision of Vegetable Commodities Based on Random Forest and ARIMA Models

Due to the short freshness period of vegetables in fresh food supermarkets, if they are not sold or replenished in a timely manner, it will lead to operational problems such as the accumulation of substandard products and oversupply. In order to explore the correlation relationship and related laws of vegetable sales, and make replenishment and pricing strategy for the next sales based on this, this paper takes the measured data of a supermarket as an example, and establishes a random forest model with the sales volume, sales price and wholesale price as the indexes, and solves the sales law of vegetables, and optimizes the strategy of replenishment and pricing of vegetables on the basis of the model. First, the Pearson correlation coefficient was used to calculate the correlation coefficient between categories. Then, the total daily replenishment quantity and pricing strategy of vegetables for the coming week were formulated, and the ARIMA model was used to predict the wholesale prices of vegetables in each category for the next seven days, and then the predicted wholesale prices were used to establish a random forest model to make sales forecasts to formulate the replenishment quantity and pricing strategy. Finally, weighted selling price, individual item code, weighted wholesale price, weekday vs. non-weekday, holiday vs. non-holiday, and seasonal information are used as indicators to train the random forest regressor, and then the ARIMA model prediction is used again to make the prediction of the wholesale price, and the optimal combination of the individual items is selected to formulate the revenue maximization strategy. The proposed model has high computational efficiency and relatively stable prediction ability.