A novel music recommender by discovering preferable perceptual-patterns from music pieces

Nowadays, advanced information and communication technologies ease the access of music pieces. However, it is still hard for the users to find what she/he prefers from a huge amount of music works. To solve this problem, most music recommenders based on collaborative filtering (called CF) utilize the rating logs to predict the user's preference. Unfortunately, CF-like recommenders cannot capture the user's preference effectively due to the gap between the complicated musical contents and diverse user preferences. To reduce the gap, in this paper, we propose a novel recommender that integrates musical contents mining and collaborative filtering to achieve high-quality music recommendation. For musical contents mining, the proposed perceptual patterns derived by Two-stage clustering are adopted as a kind of musical genes to support music recommendation. For collaborative filtering, pattern-based preference prediction can imply the user's preferred music effectively. The experimental results reveal that our proposed recommender well outperforms the existing recommenders in terms of recommendation quality.