AN ALGORITHM FOR EFFECTIVE CONTEXT-SENSITIVE MUSIC RECOMMENDATION

Mammadli Ramil Elmanoglu

Due to the growth of Internet capabilities, users are faced with difficulties in choosing music based on their worldview. Context-aware recommendation systems are widely used for music selection. To improve the capabilities of recommendation systems, important parameters are user reactions and duration of music playback. User reactions as a context have been little studied due to the difficulty of extracting feedback from user posts about specific music. The article proposes an algorithm for extracting user reactions from posts during different time intervals and presented at different levels of detail. Each piece of music obtained from the online repository VKontakte Music is presented in a triple format: (Music_id, User_id, Likes). This tuple is considered to develop several methods for providing music recommendations. Experimental tests show that the proposed method improves the performance of recommendations in terms of accuracy rate.

Keywords:User reaction; Recommendation systems; Data analysis; Data filtering