Clustering of Pharmaceutical products using Random Forest algorithm

Abstract

The purpose of this work is to uncover the patterns of products' demand variation for the case of a pharmaceutical company in order to customize the safety stock policy. It is important, for decision makers, to customize the supply chain strategies in order to match the profile of the demand in each product category. In this article, a general overview of cluster analysis and some key methods are presented. The selected algorithm for our data clustering is Random Forest. The data used for this study are related to the sales actuals of the entire portfolio of the company over a time span of five years, including some qualitative and quantitative parameters such as price, reimbursement by health insurance and type of the product. The results show four different clusters. The validity of this model has been discussed with business, medical and pharmaceutical experts.