Commodity Sales Forecast Based on Cluster Analysis and Time Series

Abstract

The price of vegetables in fresh supermarkets is affected by many factors such as time and sales. Reliable market demand analysis is especially important for replenishment and pricing decisions. This article delves into related issues in vegetable pricing. We select the sales data of the latest 2023 Mathematical Modeling Competition Question C as the data in this paper, first preprocess the data, summarize the data according to vegetable categories and single products every day, and observe the sales trend of categories and single products over time by drawing a line chart. Second, the distribution of the data is tested, and the P-P plot of each variable is drawn to determine whether it follows a normal distribution; Pearson correlation analysis and Spearman correlation analysis were used to calculate the correlation coefficient between categories and items, respectively. According to the correlation coefficient test, the model is not significant. Therefore, cluster analysis is used to explore the correlation between each vegetable category and a single product, and the cluster analysis lineage map is drawn to visually represent the correlation between each category and a single product. The data is then aggregated to plot a scatter plot between sales and cost plus price to observe possible functional relationships between them. Linear regression, loglinear regression and polynomial regression were used to train the data, plot the function curve and data scatter plot, and calculate the evaluation indicators of the regression model. The time series model is used to predict the sales volume of each category in the next week, and the relationship between sales volume and cost plus price is trained into the time series model as a penalty function. Finally, get the best replenishment volumes and pricing for each category next week so that merchants can get the highest profits.