Machine Learning Framework for Retail Sales Forecasting

Sales forecasting plays a critical role in retail operations, directly impacting inventory management, supply chain planning, and revenue optimization. This paper presents a comprehensive machine learning–based forecasting framework developed for SuperKart, a multi-city retail chain. The proposed approach integrates structured data preprocessing, exploratory data analysis (EDA), feature engineering, and comparative evaluation of regression models including Decision Tree, Random Forest, Gradient Boosting, and XGBoost. Among the tested algorithms, XGBoost outperformed others, achieving an R² score of 0.87 and demonstrating strong generalization on test data. A low-code deployment strategy was implemented using Streamlit and Hugging Face Spaces, enabling business users to generate real-time sales forecasts with minimal technical intervention. This research contributes a scalable, accessible, and data-driven forecasting solution for retail environments, bridging the gap between advanced machine learning techniques and practical business adoption.