Sistem Rekomendasi Musik Berdasarkan Playlist dengan Collaborative Filtering

Jurnal Informatika Dan Tekonologi Komputer
Vol. 5 No. 3 (2025): November : Jurnal Informatika dan Tekonologi Komputer

Modern music streaming platforms offer millions of songs, creating a challenge for users in discovering content that matches their tastes. This research addresses this problem by designing a music recommendation system using a Hybrid Collaborative Filtering approach. This method combines the strengths of Item-Based (track similarity) and User-Based (playlist similarity) filtering for higher accuracy. Utilizing 100,000 playlists from Spotify's Million Playlist Dataset (MPD), the system was developed through data preprocessing, cosine similarity calculation, and weighted score combination. The evaluation was designed using metrics like Precision@K and Hit Ratio. The results demonstrate that the hybrid model can provide thematically relevant song recommendations based on an input playlist, proving its effectiveness in personalizing the music discovery experience for users.