Design And Implementation Of A Personalized Music Recommendation System Based On Deep Learning

For music lovers,as the song library becomes larger and larger,song resources become more and more abundant,users need to spend a certain amount of time and cost to find music in line with their own interests.Usually,the method of music platform is to provide users with search function.People can only search music by song name,singer and other keywords,but this method does not consider that users are different individuals and have different tastes.One of the main goals of personalized music recommendation system is how to rightly recommend songs in line with users' interests in the huge music library.Using effective recommendation engine is an effective way to improve the personalized ability of personalized music system.This paper studies the development of music recommendation system at home and abroad,designs the important music classification model curnn based on convolutional neural network and recurrent neural network,proposes a hybrid recommendation model combining collaborative filtering recommendation model and important music attribute recommendation model,and applies the hybrid recommendation model to this recommendation system.The contributions of this paper mainly include the following three aspects: firstly,the music classification method based on neural network is studied,and on this basis,the network is improved and optimized,and a music classification model based on neural network is designed: Joint convolution neural network and recurrent neural network model curnn in series mode.Curnn model can mine more complex temporal and spatial features than CNN model,and has less parameters than RNN model.Curnn model not only can extract temporal features,but also has lighter weight and higher performance,which can more accurately use song classification for subsequent recommendation model;Secondly,in the case that the score data of Netease cloud music users cannot be obtained directly,an algorithm is proposed to establish the score matrix of music users by using the implicit data of users' playing history for subsequent recommendation;Finally,a hybrid recommendation model combining collaborative filtering and important music attribute recommendation is designed.The hybrid recommendation model increases the weight of collaborative filtering model when the user has rich listening history,and increases the weight of music attribute model when the user has less listening history.It has high flexibility in different music recommendation scenarios.Finally,this paper analyzes the feasibility and requirements of the personalized music recommendation system,designs the overall architecture of the system,and realizes the personalized music recommendation system.