Sales Prediction of Walmart Sales Based on OLS, Random Forest, and XGBoost Models

Abstract

The technique of estimating future sales levels for a good or service is known as sales forecasting. The corresponding forecasting methods range from initially qualitative analysis to later time series methods, regression analysis and econometric models, as well as machine learning methods that have emerged in recent decades. This paper compares the different performances of OLS, Random Forest and XGBoost machine learning models in predicting the sales of Walmart stores. According to the analysis, XGBoost model has the best sales forecasting ability. In the case of logarithmic sales, R2 of the XGBoost model is as high as 0.984, while MSE and MAE are only 0.065 and 0.124, respectively. The XGBoost model is therefore an option when making sales forecasts. These results compare different types of models, find out the best prediction model, and provide suggestions for future prediction model selection.