Data-Driven Machine Learning Approaches to Demand Forecasting in Supply Chain Management

Abstract

This study explores pharmaceutical demand forecasting by integrating statistical analysis and machine learning within the context of supply chain management. The dataset, from a pharmaceutical manufacturer, includes six drug categories and around 240 products. Initial data exploration was conducted using Power BI to understand the structure and trends. Statistical models, namely Moving Averages, Single Exponential Smoothing, Trend Analysis, and ARIMA, were applied, focusing specifically on Acantaine (Antibiotics class) for detailed analysis. Machine learning models were trained on the full dataset to improve generalization, but testing was focused on Acantaine, with accuracy metrics computed. A comparative analysis showed that machine learning models, especially Random Forest, significantly outperformed traditional statistical methods in forecasting accuracy. The study underscores the value of machine learning in enhancing prediction and supporting strategic decisions in pharmaceutical supply chains.