Research on Commodity Sales Forecasting Based on Stacking Ensemble Learning
Improving the accuracy of sales forecasting can assist e-commerce enterprises in more effectively planning inventory,thereby enhancing the efficiency of supply chain management.This article employs the Stacking ensemble learning method to integrate multiple models for sales forecasting.The research focuses on the selection of base learners,using RF,SVM,XGBoost,and LSTM algorithms as the first-layer models,with linear regression as the second-layer model.Through the validation with historical sales datasets,the study reveals that,compared to individual models,the Stacking ensemble learning method produces superior predictive performance.This research approach can effectively combine the strengths of various models,and yield more accurate sales forecasting so as to provide a basis for e-commerce enterprises to formulate inventory and production plans.
NETL
NSTL
万方数据
