Abstract

Sales forecasting is an essential task for effective business planning and implementation, and also inventory management. Traditional forecasting methods usually fail to identify complex sales patterns, resulting in inaccurate predictions. Therefore, this study examines the accuracy of sales prediction using machine learning models. By leveraging historical sales data, this study estimates multiple machine learning algorithms, such as random forest, neural networks, linear regression, XGBoost, and decision trees. The findings reveal that, while Linear Regression recorded high error rates, MAE (136.07%), RMSE (303.85%), MAPE (302.66, %), Random Forest achieved the lowest errors, MAE (110.10%), RMSE (234.74%), and MAPE (38.61%), showing its outstanding forecasting performance. With a better predictive accuracy, Random Forest gives businesses better insights for decision-making concerning inventory management, efficiency of operations, and meeting market demands. The study emphasizes the increase of data-driven strategies and business analytics in sustaining competitive advantage.

Keywords: Decision trees, Machine Learning, Predictive Analytics, Regression models, Sales Forecasting