An Ensemble Kernel Gaussian Process Regression Framework for Predicting OTC Pharmaceutical Sales: An Empirical Analysis on Multinational Transaction Data

Abstract

This study aims to enhance the predictive accuracy of over-the-counter (OTC) pharmaceutical sales by applying an Ensemble Kernel Gaussian Process Regression (EK-GPR) model to a multinational transaction dataset from January to August 2022. The analytical workflow included data cleaning, temporal feature engineering, categorical encoding, and numerical normalization to obtain an optimal representation for kernel-based modeling. The results demonstrate that the EK-GPR model—combining RBF, Matern, and Rational Quadratic kernels—achieves superior predictive performance compared to single-kernel GPR and benchmark models such as Random Forest and XGBoost. Residual analysis confirms that EK-GPR effectively captures complex non-linearities and seasonal variations while providing stable uncertainty estimates that support risk-informed decision-making. These findings highlight the potential of EK-GPR as a robust and effective forecasting approach for pharmaceutical demand planning, with important implications for inventory management, supply chain optimization, and the development of real-time predictive systems.