An Improved Rating Mapping Algorithm for Music Recommender System

Existing rating mapping algorithm maps each user's rating for a particular artist based on the complementary cumulative distribution interval where the artist play count lies. When the artist play count lies across two intervals, the higher rating is not appropriate when the majority of artist play count lies in the lower interval of the distribution. To solve this problem, we use the median of the artist play count instead as the statistical object to optimize existing algorithm. Experimental results show that the improved method can obtain a general increase in recommendation accuracy by about 1% compared to the existing algorithm.