Prediction Of Supermarket Merchandise Sales Based On Data Mining

With the development of Internet technology and the widespread use of big data,online and offline shopping methods have brought consumers a more convenient consumer experience.But at the same time,for the physical supermarkets,the competition they face is even fiercer.How to deal with the problem of consumer churn,adopt a sales policy that more caters to consumers' psychology,and increase business competitiveness,which is a fundamental and fundamental issue for a company.This article is aimed at a supermarket chain,using data mining technology,based on the relevant data currently available,designing relevant algorithms,and forecasting the sales of goods in the next month.Through accurate forecasting,not only can the company managers provide relevant decision support,but at the same time they can provide technical support for the next development of the company.This article first introduces the current status of data mining technology in retailing commodity forecasting,describes the main content of this article,the overall route and research methods,and proposes a research based on data mining technology for supermarket commodity sales forecasting.This article focuses on how to use data mining for short-term sales forecast research.First of all,this paper conducts a visual analysis of the data and performs data preprocessing operations on the discovered abnormal data.Through the introduction of feature engineering construction,this paper focuses on the construction of feature engineering from various dimensions and fully exploits the characteristics of the relevant dimensions.In the model construction,this paper first carries out the multiple linear regression prediction model,the support vector regression prediction model,the random forest regression prediction model,the gradient descent regression prediction model based on the model prediction,and will use the support vector regression model with better results and the main The component analysis method and the clustering algorithm were combined to construct the dynamic feature vector extraction,and the basic sales factor was obtained.Then,for different training sets,LightGBM was used for the final result prediction.The experimental results show that the model has an ideal forecast result for the short-term commodity sales forecast.