Hierarchical clustering for collaborative filtering recommender systems

Abstract

Nowadays, the Recommender Systems (RS) that use Collaborative Filtering (CF) are objects of interest and development. CF allows RS to have a scalable filtering, vary metrics to determine the similarity between users and obtain very precise recommendations when using dispersed data. This paper proposes an RS based in Agglomerative Hierarchical Clustering (HAC) for CF. The databases used for the experiments are released and of high dispersion. We used five HAC methods in order to identify which method provides the best results, we also analyzed similarity metrics such as Pearson Correlation (PC) and Jaccard Mean Square Difference (JMSD) versus Euclidean distance. Finally, we evaluated the results of the proposed algorithm through precision, recall and accuracy.