Music recommendation based on content and collaborative approach & reducing cold start problem

Abstract

Nowadays, Recommendation becomes the most popular area for many researchers. The main aim of recommendation is to provide meaningful suggestions to users for particular item based on users past interest and behaviors towards items. There are two most popular recommendation algorithm is 1) Content-Based Filtering 2) Collaborative Filtering. Content-Based method recommends music based on user data. Collaborative method uses rating and content sharing between different users to recommend music. Here, to provide music recommendation by content-based method music subjective features Speechiness, loudness, Acoustiness etc. are analyzed. The extracted features are stores into database by using Kmean clustering algorithm. For Content-based method, whenever user fires query to database music feature attribute value compares with clusters centroid. Once attribute value match, music can be recommended to user as Content-based method. For collaborative method, rating given by user to particular music is considered and adjusted cosine similarity is used to find similarity between user-user. Once similarity found, prediction rating algorithm is used to provide recommendation to user. Cold-start is most common problem for new user. Here, most popular tracks are recommending to user to solve it.