AI-Driven Demand Sensing and Forecasting in the Face of Volatile Global Markets

Abstract

Accurate demand forecasting is essential for effective supply chain management, particularly in volatile and uncertain market environments. However, traditional forecasting techniques often struggle to capture nonlinear demand patterns and sudden market fluctuations, resulting in operational inefficiencies such as inventory shortages or excess stock. This study investigates the effectiveness of artificial intelligence (AI)–driven demand sensing and forecasting, with a particular focus on Long Short-Term Memory (LSTM) neural networks, in comparison with conventional statistical forecasting models. A simulation-based experimental design was employed to generate demand data characterized by high volatility, enabling a controlled evaluation of forecasting performance across multiple models. Forecast accuracy was assessed using standard error metrics alongside visual trend analysis. The findings reveal that the LSTM model consistently outperforms traditional approaches by closely tracking actual demand movements, exhibiting lower forecast error variance, and producing fewer extreme prediction errors. These results underscore the capability of AI-based forecasting systems to improve supply chain responsiveness and resilience in uncertain and rapidly changing market conditions. This study contributes to the existing literature on AI applications in operations management by providing empirical evidence supporting the superiority of deep learning techniques for demand sensing and forecasting in volatile global markets.