Demand Forecasting for Unpredictable Environments: Are We Ready?

Leveraging data and technology can significantly improve the retail sector's efficiency, customer experience, and business performance through advanced forecasting models. In this study, the critical role of accurate demand forecasting is emphasised in optimising resource allocation, inventory management, and strategic decision-making in retail operations. In the context of demand forecasting and time series analysis, this study introduces a multi-layer long short-term memory (LSTM) network for predicting future demand, showcasing its superiority over traditional methods like ARIMA, SARIMA, and Prophet. Through a comparative analysis, the LSTM-based approach demonstrates enhanced performance in forecasting demand for goods in an automotive retail setting. The proposed LSTM presents lower Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Symmetrical Mean Absolute Percentage Error (SMAPE) offering a maximum reduction of 72%, 84%, and 69% respectively.