Performance Comparison of Collaborative Filtering Using Fuzzy Clustering for Spherical Data
Tadafumi Kondo, Yuchi Kanzawa · 2018
GroupLens is the representative neighborhood-based algorithm for collaborative filtering task, where the definition of "neighborhood" is heuristic. This study proposes determining the neighborhood using fuzzy clustering for spherical data, and compares three fuzzy clustering for spherical data been proposed with GroupLens algorithm, using two real datasets. The experimental result shows that the proposal achieved higher recommendation accuracy for all two datasetsand that all three clustering algorithms help determining more adequate neighbor-hood than the conventionally determined neighborhood.