Enhancing FMCG sales prediction using recurrent neural networks and economic indicators

Sales forecasting is key to maximizing inventory control, manufacturing planning, and strategic planning within the Fast-Moving Consumer Goods (FMCG) industry. This study suggests a prediction model for FMCG product sales forecasting based on economic indicators and past sales (lagged values) with the help of a Recurrent Neural Network (RNN). To identify temporal relationships and external economic consequences, the model employs the inflation rate and past sales (lagged values) as leading predictors. Mean Absolute Percentage Error (MAPE), Root Mean Squared Error (RMSE), and R2 score were employed to measure the performance of the model. The result indicates that the model possesses a low MAPE of 1.01%, indicating high prediction accuracy 98.99%, and R2 score of 0.9837, which explains 98% of sales data variance. The credibility of the model is also established through residual analysis and correlation findings, indicating that historical sales possess a high correlation with future demand and that inflation negatively affects sales performance. Although the RNN model performs well, it can be enhanced by incorporating seasonality, other macroeconomic variables, and sophisticated topologies such as Long Short-Term Memory (LSTM) networks for more in-depth sequential learning. The outcomes indicate the accuracy of deep learning in forecasting FMCG sales, which may assist companies in making data-driven marketing and inventory decisions.