OP-K-Means: Optimized Algorithm for Recommendation System Based on User Preferences

Abstract

Abstract For a recommender system (RS), it is difficult to capture all the user’s interest lists simultaneously, which leads to the problem of insufficient performance of the existing joint RS based on the K-Means clustering algorithm. In this paper (1), we introduce a cluster optimization method OP -K-means for user preference data. This method starts with propagation from the center of the user preference data. By selecting relatively distant positions between each initial center, the distance between them is increased as much as possible. (2) Finally, we validate the effectiveness of our algorithm on a dataset from Facebook and compare our algorithm with original K-means. Our experimental results justify the validity of our OP -K-means algorithm.