Enhancing Retail Sales Predictions for Big Mart Using Advanced Machine Learning Techniques

Abstract

In today's competitive retail environment, accurate sales forecasting is essential for efficient inventory management and
maximizing profitability. Supermarket chains like Big Mart generate vast amounts of sales data, which, when analyzed effectively, can
provide valuable insights into future demand. This study presents a predictive analysis framework using advanced machine learning
algorithms to forecast sales for Big Mart stores. The proposed system implements multiple algorithms, including XGBoost, Linear
Regression, Polynomial Regression, and Ridge Regression, to evaluate sales patterns and trends. By applying these techniques to ten years of historical sales data, the study compares model performances based on metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). Results indicate that the XGBoost model consistently outperforms other models in terms of accuracy and predictive capability. Furthermore, the system addresses challenges like data anomalies and multicollinearity using Ridge Regression for robust predictions. The insights generated can assist retailers in decision-making processes, enabling better inventory planning, demand forecasting, and strategic resource allocation. Future enhancements may involve incorporating external factors such as economic indicators and customer sentiment analysis to further refine prediction accuracy.