Sales forecasting and classification model and pricing research of agricultural perishables based on STL-ARIMA

Abstract

The prediction of sales volume for agricultural perishable goods is crucial for small-scale merchants to formulate corresponding procurement strategies. Accurate sales volume forecasting helps merchants reduce the risk of excess inventory, establish prudent and appropriate procurement strategies, and provide suggestions for product pricing. To further accurately explore the sales volume patterns of various perishable goods, this study classifies existing data into individual products using the STL (Seasonal-Trend decomposition using Loess) algorithm, revealing the characteristics of each product's sales volume over time. Based on the existing sales volume data, the ARIMA model is used to predict the sales volume of specific categories of goods. A multivariate linear regression model related to pricing, based on existing cost and pricing data, is established to provide merchants with pricing suggestions.