Forecasting Product Demand Using The N-Beats Model In Retail Management

Abstract

Demand forecasting is important for proper inventory control and business operation efficiency in the retail industry. Nevertheless, classical methods of forecasting, such as the ARIMA method, are not capable of forecasting demand accurately due to their inability to cope with nonlinearity in the data. This is the reason why this paper seeks to examine the use of the N-BEATS method in demand forecasting within the retail sector. The research objective will assess the effectiveness of N-BEATS method in demand prediction compared with other forecasting methods. In this research, a retail sales dataset comprising product sales data, promotional events, and exogenous variables was employed. Preprocessing of the Data was carried out before training the N-BEATS model. Comparison of findings with existing theories is done by looking at the accuracy and forecasting potential of the developed approach. In this regard, the N-BEATS model was found to perform better compared to the rest of the models, since it had the least RMSE (0.895 ± 0.004) and MAPE (18.5% ± 0.1%). This shows that the N-BEATS model is better at forecasting the demand for products. Models such as ARIMA and LSTM recorded high error rates, thus depicting the ability of N-BEATS model to manage complex demand forecasts. N-BEATS model is an excellent way over which demand forecasting can be done in retail outlets.