Music Recommendation System based on Listening History Data

The ease of accessing music as well as the fast growing numbers of new musical pieces has made music listeners spend more time to choose music pieces in order to make a suitable playlist. Music recommendation can help music listeners to make a suitable playlist with just a minimum effort and time. There are various music features that have been used to produce music recommendations, such as music content, music context, user properties, and user context. In the current research, context features that are more related to user properties will be used to create a playlist of music recommendations. The features used are demographic features such as age, gender, and country. Listening history from users are also collected, so it is known what type of music is often listened to at certain times. These features are expected to make music playlists more user-friendly when compared to only using music content features. The Self Organizing Map method will be used to classify the listening history data. Music in the same cluster (having many similarities) will have a higher chance being in the same playlist. The recommendation system built in this research has an average precision of 0.606. The precision value obtained is not high, it is necessary to add a recommendation feature that is closer to each user's personalities, such as musical genre preferences to increase the value of precision.