Application of recurrent neural networks in demand forecasting: A teaching brief

What the paper says

Abstract Accurate demand forecasting is critical for optimizing supply chains, yet traditional statistical models struggle with complex, nonlinear demand patterns. Recurrent Neural Networks (RNNs) offer a powerful alternative by capturing sequential dependencies and adapting to shifting trends. However, RNNs steep learning curve presents challenges for business students with limited technical backgrounds and instructors with varying technical expertise. This study responds to such challenges by developing a structured instructional framework that bridges the gap between traditional forecasting and RNN‐based techniques in business education. Through a scaffolded learning approach with comprehensive implementation support, students progressively transition from statistical models to hands‐on RNN implementation using Python. Pre‐module and post‐module assessments demonstrate significant gains in conceptual understanding, technical proficiency, and the ability to apply RNN‐driven forecasts in decision‐making contexts. A high level of student engagement further reinforces the effectiveness of this approach. This study contributes to the literature by advancing RNN education beyond purely technical considerations through the development of a structured instructional framework accessible to business students with limited programming experience and instructors across diverse institutional contexts. By establishing an adaptable educational model, this study facilitates the integration of RNN‐driven forecasting into business curricula, enhancing students' ability to apply machine learning techniques in real‐world contexts.