Content-Based Music Recommender System Using Deep Learning

With the rapid development of the music streaming media service industry,users can easily hear any song on mobile devices,and the Internet has become a huge music storage platform.At the same time,how to find favorite songs from massive data has become a very tricky issue.The music recommender system is the best way to solve this problem.It can recommend the user’s favorite songs to the user,and it can also find the right target users for the songs.Music recommendation is an important application area of recommender system technology.Due to the particularity of the music recommendation field,the most effective collaborative filtering method is not suitable for the music recommendation field.The reason for this is on the one hand because the music field does not have enough user rating data,on the other hand,the collaborative filtering method leads to cold start,so that new songs cannot be recommended.An effective way to overcome these two problems is to build a content-based music recommender system.This paper mainly studies the content-based music recommendation system,which includes two important processes of music audio feature extraction and recommending songs to users.This paper revolves around these two processes to improve and optimize,so as to build a more effective content-based music recommendation system.The specific research content is as follows:1)In order to extract a better musical audio feature representation.We propose a model that mixes two different types of deep neural network structures to extract the audio features of music.Different deep neural network structures are suitable for extracting different aspects of features.The hybrid deep neural network model can extract the effective feature representation of songs in classification tasks of different classification criteria and achieve better classification result.2)We uses the method of matching user profile and music audio features as our recommended strategy,and tries to predict the probability that the user is interested in the songs.We extracted the features of each song through the hybrid deep neural network model,and obtained an accurate user profile through a series of feature engineering.Machine learning method was used to learn the user’s interest in each song.The experimental results show that our recommender system can produce effective recommendations both subjectively and objectively.