Demand Prediction Based on Machine Learning Algorithms for Optimal Distribution of Insulin

Abstract

The pharmaceutical supply chain is a complex network involving a wide range of stakeholders and requiring many steps to ensure the availability and right distribution of drugs. So, good demand prediction is important to improve the satisfaction of patients' needs. Thus, this study proposes a demand prediction based on machine learning algorithms in order to forecast the required quantity of insulin in the USA. Our study concentrated on the application of some machine learning models: random forest, multiple regression, and artificial neural network. We compare the performance of these three models based on the regression metrics: the root mean squared error, the coefficient of determination and the R-value, in order to choose the best model that is appropriate for our research. After that, the value of the uncertain demand predicted by the best proposed machine learning model will be used in a mathematical model, which focuses on the downstream part of pharmaceutical supply chain between the distributor and the demand zones, and it aims to maximize customer satisfaction.