Optimized similarity computation for nearest neighbor choosing

Abstract

Finding appropriate nearest neighbors is the essential task to User-based collaborative filtering system. And computation for getting the similarity of different users is the key step of locating those nearest neighbors. Existed similarity computation algorithms adopt various filling methods to achieve better performance in circumstance with disperse dispersed data set, meanwhile - however - these filling methods bring inaccuracy into similarity computation. For resolving this, this paper proposes a optimized similarity computation algorithm that eliminate those inaccurate factors-items only be rated by one of two users- from conventional similarity computation.