IMPLEMENTASI METODE HYBRID COLLABORATIVE FILTERING DAN CONTENT-BASED FILTERING UNTUK REKOMENDASI LAGU

Abstract

The goal of this study is to develop a song recommendation system using hybrid filtering method by combining collaborative filtering and content-based filtering method. Collaborative filtering recommends songs based on the similarity of users’ tastes. Content-based filtering provide song recommendations based on the similarities in their audio characteristics, such as genre, tempo, and pitch. Hybrid filtering is a recommendation method that combines two or more methods.  The data that are used to build the recommendation system are users' song playback history and audio feature data from the songs. Test results show that the collaborative filtering method has the highest F1-Score value, 0.0713, while the content-based filtering method has the lowest F1-Score value, 0.0060. The weighted hybrid filtering method has an F1-Score value of 0.0550 and the highest hit-rate compared to other methods, 0.7137.