Research On Music Recommendation Algorithm Based On Collaborative Filtering

With the rapid development of the network information age,the degree of information redundancy is deepening day by day.In today's society,all kinds of resources are not only rich in categories,content is also very complex,so that users can not efficiently and accurately filter out information in a short period of time to meet their own requirements,the emergence of recommended algorithms to a large extent effectively alleviate this problem.The recommended algorithm has been applied to more and more fields,especially in e-commerce.At the same time,the application of recommendation algorithms in entertainment has become more and more common,such as music recommendations and movie recommendations.More and more people take listening to music as a way to relax after work and study,however,there is a serious problem of information asymmetry between users and music,how to obtain songs in line with the user's preferences in the huge amount of music has become worth thinking about.Collaborative filtering algorithms stand out among many algorithms with their concise,easy-to-implement and good recommendations,gain the unanimous approval of experts and developers,and then begin to apply people's learning and life on a large scale.The collaborative filtering algorithm is generally divided into two categories: the user-based co-filtering algorithm and the item-based collaborative filtering recommendation algorithm.The difference between the two algorithms is that the dependent objects are different.However,the traditional collaborative filtering recommendation algorithm has certain limitations,just by analyzing the user's rating of the item to determine the user's preferences,ignoring the important influence of the tag this element,such a way is neither objective nor accurate,can not be targeted to collect user interests.In addition,there is a problem of sparse user data producing cold start,can not provide users with a quality service experience.In view of the above problems,this paper creatively studies and puts forward a label-based collaborative filtering recommendation algorithm based on time factors,and label-based collaborative filtering recommendation algorithm that integrates user interest.The user's preferences are obtained by building a user scoring matrix,which is then combined with the label weight calculated by the TF-IDF method and a model of the distribution of user interests is constructed.On this basis,the fusion time factor and user interest degree,the user similarity model is constructed,through the experimental test of last.fm music data set,the accuracy,recall rate and F1 value of the algorithm are obtained from the traditional algorithm and the proposed algorithm,which proves that the newly proposed collaborative filtering algorithm of fusion time factor and user interest is more accurate and better than the recommended results of the traditional collaborative filtering algorithm.