Research And Implementation Of Personalized Music Recommendation System Based On Mixed Model

With the development of Internet, the amount of information is increasing, and how to find the content we really want in a large amount of information has become the focus of the current study. Traditional search engines require users to be able to accurately describe what you want to search, but sometimes the user is not well described or they do not know what to want. The emergence of the recommendation system for users to solve this difficulty, it does not require users to enter a large number of search keywords, but through the analysis of the user’s historical records, to give users a personalized recommendation.Music is a very suitable for the recommended content, everyone has their own music taste, recommend the system for each user to customize the personalized. Compared with representative music recommendation system and related algorithms, analyzes the advantages and disadvantages of each recommendation algorithm, then proposed based on collaborative filtering and music gene of mixed mode is recommended. Through the experiment compared based on Collaborative Filtering Recommendation Based on gene of music recommendation and recommendation based on the hybrid model, and finally selects recommendation based on mixed mode as the music recommendation system recommendation algorithm is presented in this paper.In collaborative filtering based recommendation model, we first construct the user item matrix, and then analyze the similarity of users, find the nearest neighbor users, and get a list of recommendations. In the recommendation model based on the music gene, by analyzing the structure of music gene, the gene preference of several users is selected and calculated to generate a recommendation list. In the mixed recommendation model, a final recommendation list is obtained by weighting the above two.Based on Collaborative Filtering Recommendation application range compared to other recommended more widely, the recommendation accuracy is relatively high, but there are also some sparsity and cold start problems; based on the recommended content without relying on user evaluation information. At the same time, there is no association with the same filter of sparsity and cold start disadvantages. But the application scope is relatively smaller. Based on the hybrid model of the recommendation for these two shortcomings, can take different weights in different situations. When the user’s evaluation information is relatively small, increase the weight based on the recommendation of the music gene; when the user’s evaluation information is more, increase the weight based on collaborative filtering recommendation. In this way, the advantages of collaborative filtering based on collaborative filtering and music gene recommendation are comprehensively utilized.This paper designed and implemented a personalized music recommendation system, which is a B/S architecture of the Web music website, the use of the current popular CodeIgniter and Bootstrap framework, and finally the system was tested.