Optimization XGBoost Algorithm Using Parameter Tunning in Retail Sales Prediction

In retail companies, the owner needs sales analysis to make decisions in the company's business processes. Several previous studies have introduced forecasting techniques using regression analysis, and classification approaches that need optimization. This article proposes a new approach to sales prediction using XGBoost, which is optimized by comparing the best performance from three optimization methods: Random search, grid search, and Bayesian optimization. The aim is to obtain the best comparative analysis and increase prediction accuracy. The novelty of the proposed model is determining the best value for each optimization method using XGBoost. The results of the evaluation show that the best results were achieved by the grid search optimization technique in the XGBoost model with an increase in the evaluation value R^2 from 97.31 to 98.41. The results of the proposed model analysis can help retail business owners in accurate sales predictions to determine the development of business processes.