Expiry-Based Dynamic Discount System: A Machine Learning Approach to Perishable Inventory Pricing

Abstract - In the retail and Fast-Moving Consumer Goods industries, perishable goods such as dairy products, bakery items, fresh produce, and pharmaceuticals face persistent challenges in inventory management and revenue optimization. Products approaching expiry often remain unsold, resulting in financial losses and environmental waste. Traditional discount strategies are static, manual, and lack data-driven intelligence. This paper presents an Expiry-Based Dynamic Discount System that leverages machine learning and data analytics to automatically compute optimal discount percentages based on product freshness, days to expiry, overstock ratios, and category-specific parameters. The system employs Ridge and Lasso regression models with hyperparameter tuning, combined with a formula-based feature engineering pipeline including freshness scores, urgency levels, and perishability assessments. An interactive Streamlit dashboard enables store managers to upload inventory, apply filters, override discounts, and export recommendations. Experimental results demonstrate a projected 67% reduction in product waste and a 12% net revenue improvement over manual approaches, with regression models achieving R² values above 0.85. Key Words: Dynamic Pricing, Machine Learning, Perishable Inventory, XGBoost, Gradient Boosting, Retail Analytics.