Demand Forecasting in Retail Using Multi-Channel Consumer Behavior Data

The accuracy of demand forecasting remains a critical factor in retail supply chain optimization. With the proliferation of digital platforms, multi-channel consumer behavior data presents a novel opportunity to improve forecasting precision. This study proposes and evaluates advanced machine learning models for retail demand forecasting using data collected from online and offline customer interactions, including purchase history, web traffic, mobile app activity, and in-store transactions. Predictive models such as Random Forest, XGBoost, and Long Short-Term Memory (LSTM) neural networks are deployed. Evaluation is conducted using root mean square error (RMSE), mean absolute error (MAE), and R-squared (R²). The results reveal that LSTM outperforms conventional models, particularly in capturing temporal demand shifts and promotion-driven surges. The findings highlight the importance of integrating heterogeneous behavioral data and using advanced learning algorithms to inform inventory planning and marketing strategies. Keywords: Retail Analytics, Demand Forecasting, Consumer Behavior, Multi-Channel Data, Machine Learning, Time Series Forecasting