Prediction Of Merchandise Sales Based On Improved Decision Tree Model

For the sale of commodities,the most important thing is sales.In order to effectively reduce business losses and improve the profitability of shopping malls,it is necessary to make an accurate forecast of the sales situation of shopping malls.In view of the characteristics of sales data in large shopping malls,this paper analyses the factors that may affect the sales of commodities themselves and the external environment.Statistical theory knowledge including supervised learning,model selection and evaluation,error and over-fitting is expounded.On this basis,the widely used decision tree model is expounded.Based on the decision tree model,in order to reduce over-fitting and improve the generalization ability of the model,a gradient lifting decision tree model based on decision tree and integration idea is established,and compared with the BP neural network model.In order to overcome the shortcomings of using only the first derivative in the gradient lifting decision tree model and avoid over-fitting,a better XGBoost model is established.This paper forecasts the sales data from 2013.By establishing feature engineering,a prediction model is established.The experimental results show that the gradient lifting decision tree model and the BP neural network model are better than the decision tree model,but the running time is longer.XG-Boost model not only prevents over-fitting,but also improves the running speed.Therefore,the XGBoost model performs best.