An Intelligent Decision Support System for Retail Sales Forecasting Using Machine Learning and Explainable AI

Abstract

Accurate retail sales forecasting is essential for effective inventory management, efficient supply chain operations, and informed strategic business planning. Although machine learning techniques have significantly improved forecasting accuracy, their limited interpretability often reduces user confidence and restricts their practical adoption in business environments. This paper presents an Explainable Intelligent Decision Support System designed to bridge the gap between predictive analytics and business decision-making. The proposed framework employs an Extreme Gradient Boosting model enhanced with time-series feature engineering, including lag features, rolling statistics, and calendar-based attributes, to capture complex demand patterns. The developed model achieved a coefficient of determination of 0.9332, a mean absolute error of 6.2698, and a root mean squared error of 8.1708, demonstrating strong forecasting performance. To improve transparency, TreeSHAP is integrated to provide both global and local explanations of model predictions, enabling users to understand the factors influencing forecast outcomes. These explanations are subsequently incorporated into a Multi-Criteria Decision Analysis framework that computes decision and risk scores, assisting retailers in making informed inventory planning decisions. The complete framework is implemented as an interactive Streamlit dashboard that presents forecasts, explanations, and recommendations through a user-friendly interface.