ADVANCED DEMAND FORECASTING FOR RETAIL SUPPLY CHAIN MANAGEMENT USING DATA SCIENCE AND MACHINE LEARNING INVENTORY OPTIMIZATION

This project presents an interactive Streamlit-based web application that enables advanced demand forecasting for retail supply chain management using a pre trained XGBoost machine learning model. The main goal is to optimize inventory by accurately predicting daily item-level sales based on features like store ID, item ID, and date attributes (year, month, day, day of week). Users can upload a CSV file containing test data, and the application performs real-time sales predictions using the trained model. The app provides rich visual analytics including monthly sales trends, store and item-specific breakdowns, and actual vs. predicted comparison charts. It also highlights the month with the highest sales and calculates metrics such as Root Mean Square Error (RMSE) when actual sales data is available. Users can download the forecast results as a CSV file for further analysis. The application uses data preprocessing, Seaborn and Matplotlib for plotting, and joblib for loading the XGBoost model efficiently. Through an intuitive interface, it empowers retailers to make data-driven inventory decisions, reduce overstock or understock issues, and enhance operational efficiency