SALES FORECASTING USING MACHINE LEARNING

Abstract - Sales forecasting remains a critical component in the strategic planning and operational efficiency of retail enterprises. This study presents a data-driven forecasting model tailored for retail sales prediction, using a real-world dataset comprising historical sales records, outlet characteristics, and product features. The methodology emphasizes robust data preprocessing, including handling of missing values, outlier treatment, and categorical encoding, followed by the application of ensemble-based regression. Among various algorithms evaluated, the Extreme Gradient Boosting Random Forest Regressor (XGBRFRegressor) demonstrated superior predictive performance, achieving consistent accuracy across cross-validation folds. To enhance practical applicability, the forecasting system was integrated into an interactive interface, enabling real-time prediction based on user-defined input parameters. The proposed approach offers a scalable and reliable framework for informed decision-making in retail operations, particularly in inventory management, demand planning, and revenue optimization. Key Words: predicting, forecasting, demand planning, sales