Optimizing Inventory Management in Retail Companies Through Sales Prediction Using XGBoost

This study proposes a novel approach to optimizing inventory management in retail companies by integrating a machine learning-based demand forecasting model using the XGBoost algorithm. Unlike traditional forecasting methods, this research leverages the power of gradient boosting for enhanced predictive accuracy. A real-world dataset comprising 78,000 transaction records from January 1, 2024, to January 31, 2025, containing 10 attributes such as product ID, quantity, price, and store ID was utilized. The dataset underwent data cleaning, transformation, feature selection, and splitting into training and testing sets. The proposed model achieved a Root Mean Squared Error (RMSE) of 0.88, demonstrating strong prediction performance. The results were integrated into the Odoo inventory system to enable automated restocking. The novelty of this work lies in the weekly aggregation of retail data and the application of XGBoost for SKU-level demand prediction in real-time system deployment. Findings indicate significant potential for minimizing stockouts and overstock, improving supply chain efficiency, and supporting data-driven procurement policies.