Demand prediction model for supply chain optimization

Abstract

This study proposes a demand prediction model based on artificial intelligence to optimize the supply chain of the pharmaceutical distributor Grupo Montalvo. The main objective is to improve the accuracy of medication demand forecasting using both statistical and machine learning models, including Simple Exponential Smoothing (SES), ARIMA, SARIMA, Holt-Winters, Random Forest, neural networks (NN), and LSTM. The models were evaluated using performance metrics such as Mean Squared Error (MSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). Results showed that the neural network model (NN) performed the best, achieving an MSE of 8.10e-06, an RMSE of 0.0028, and an MAE of 0.0016. In comparison, the Random Forest model had an MSE of 2.04e+09 and an RMSE of 45232.60, indicating significantly lower performance. The LSTM model, on the other hand, displayed an extremely high MSE of 1.07e+21, suggesting the need for further adjustments. These findings highlight the superiority of neural networks in providing accurate demand predictions, contributing to more efficient supply chain decision-making and optimization.