Validation of a pre-clustering strategy for a recommender system

Abstract

In this work we used a proof of concept to validate a pre-clustering strategy using firmographics data in order to improve the performance of Neoway's recommender system. The context of recommender systems is discussed, together with an introduction of the concept of clustering and principal component analysis. The terminology of the recommender system and its benchmark are explained, along with the internal blocks of the system. Clustering procedures are discussed and two experiments are proposed. The results showed a slight improvement on the performance of the system, but it is not enough to proceed with the further investing in the approach.