Data driven Approaches for Demand Forecasting in Supply Chain for Business Decisions

Demand forecasting is viewed as a complex challenge for industries as it relies heavily on various factors. The Supply Chain Management (SCM) integrates processes vital for corporate environments across industries. This paper presents a comprehensive data-driven framework for sales demand forecasting in SCM aiming to enhance accuracy of forecasts. Using Rossmann Stores data case study the sales are predicted up to 6 weeks ahead with a comparative analysis conducted for long-term predictions. Multi-dimensional data complexity is addressed by identifying relevant features and selecting appropriate algorithms. The proposed framework utilizes four data-driven machine learning algorithms: Extreme Learning Machine (ELM), Decision Tree (DT), Random Forests (RF), and Extreme Gradient Boosting (XGBoost) aiming to construct a high-precision sales demand forecasting model. The experimental validation using Rossmann Stores data demonstrates the effectiveness of the proposed framework, highlighting its significance in improving the accuracy of forecasts in SCM and enhancing decision-making processes.