Modified Collaborative Filtering Algorithm Based on Multivariate Meta-similarity

摘要

This paper further research the recommendation algorithm bases on the meta-similarity. We consider more information about users collect the items, and define the epidemic degree of the item(EDI) and user(EDU), modify the degree of overlapping of items, and analyze the effect of multivariate similarity in the recommendation system, then we present a modified collaborative filtering algorithm based on multivariate meta-similarity (MMSCF). The method reduces the influence of the EDI and EDU, limited the error to transfer, and enhances the similarity by multivariate meta-similarity. The experiments prove the new recommendation algorithm evaluated by the precision indexes of ranking score, precision and recall have achieved significantly improve.