Application of Similarity Metrics in Collaborative Filtering Based Recommendation Systems

Abstract

This paper explores the ways in which various similarity metrics can be applied in recommendation systems in machine learning that are based on collaborative filtering. It examines properties of different similarity metrics often found in recommendation systems and presents findings of tests done on data sets of different sizes and data properties where these metrics were applied. The findings presented in this paper give guidance for the appropriate application of similarity metrics in machine learning and specifically recommendation systems based on collaborative filtering.