Design And Implementation Of Tag Based Music Recommendation System

The role of music recommendation in people’s daily life has become increasingly important. Music tags, which can reflect the music information, but are often ignored by traditional collaborative filtering music recommendation, may affect the results of recommendation. At the same time, due to the special properties of music, the user’s preference for music would change as the environment changes, making the music recommendation different from other item recommendation. Therefore,how to use the tag to mine the song information and combine with the user’s immediate status to achieve more effective recommendation is a hot topic for current scholars.Based on the traditional user-based collaborative filtering recommendation, this paper classifies the music tags and extracts the user’s immediate status from the song playlist to make music recommendation meets the users’ immediate status. The main contents are as follows:(1) This paper proposes a music tag classification method based on the LDA topic model which supply similar tag for music. It can ensure the original weight of the tag and extend the label, and the music label can be classified by this method. The experiment shows that this method can complete the music tag classification, and achieve the purpose of extracting the topic of song.(2) The user’s song sequence is transformed into a song theme sequence, and the frequent topic sub-sequences are extracted by the sequence pattern mining algorithm, which is used as the input of the sequence prediction to predict the next topic set of the user. Experiments show that the method has high prediction accuracy, and can capture the immediate status of the user listening session.(3) A music recommendation method based on the fusion of tag-topic information is proposed, which combines the previous research results with traditional user-based collaborative filtering results, and calculates the initial score and instant score of each candidate song.Re-ordering of the original recommendation list according to the score, to recommend songs for users which meets their immediate status. The experimental results show that the accuracy of the result of the song tag-topic information is obviously higher than that of the traditional method.(4) On the basis of the above three research contents, a music recommendation system is constructed based on tag-topic mining layer,song topic sequence mining and prediction layer and fusion tag-topic music recommendation layer, and a tag based music recommendation system is implemented.