Vegetable Price Forecasting Based on ARIMA Model and Random Forest Prediction

Abstract

Given the trouble that the freshness length of vegetable commodities is incredibly brief and the fine will deteriorate with the expansion in promoting time, we elevate our arithmetic processing based totally on the present records to remedy the fundamental coefficients; utilize the Pearson correlation coefficient evaluation to get the visualization of the warmness map and inside the relationship between the whole quantity of income of every class and the fee plus pricing relationship; due to a large amount of data, multi-feature modeling is proposed using random forests to merge several aspects while predicting revenues, which provides good resistance to noise for most datasets and is less likely to fall into overfitting. Then, the random forest model is applied to predict the revenue volume; the wholesale charges of saleable small products are predicted by the ARIMA model; the complete revenue corresponding to the constant revenue charge of each category is calculated; the procedure is optimized with the help of constraints; and finally, the multi-feature prediction of vegetable demand is made according to the random forest model, which in turn gives the replenishment and pricing strategy. Through the above analysis, vegetable supermarkets can make higher pricing and inventory choices to maximize profits.