Research On User Behavior Based Music Search And Recommendation

With the rapid development of the Internet,a variety of information filled with the Internet,the amount of online data is almost exponential growth.Since the stock of data is so huge and is growing rapidly,it is very difficult for users to get the small portion of information which they need,in this situation,it can lead to the problem of “information overload”.Text information is the most common form in the Internet information,it can be used to describe other forms of information,such as multimedia,pictures,etc.Sometimes,we can find the information which we truly need by querying the text information.Search engine is a tool for retrieving information,when the user submits the query keywords to the search engine,search engine begins to work,it retrieves the relevant content from the documents collection and then return the result to the user.The qualities of search results of different search engines are not the same,through the analysis of user behavior,we can understand the users' search habits,to make a better search engine which is more in line with the user needs.From the perspective of information acquisition,users can get information passivity,that is,system takes the initiative to push information to users.This technology which can provide users with information is called recommendation system.Recommendation system need to understand the user's historical behavior at first,and then try to find users who have similar historical behaviors to the user or find items which are similar to the items that the user liked through different ways,finally generate an item list for recommendation.Based on the above analysis,this paper completed the following research and work:Research on the theory and technology of search engine and recommendation system which are used in solving the problem of information overload,including the basic structure of the search engine,Chinese word segmentation technology,full-text indexing technology,Solr open source enterprise search platform,popular recommendation system techniques,and some approaches to evaluate recommendation system.Based on the user search/access log and market research,we analyzed the function requirements of the search engine for mobile music application.Then implementing a music search engine based on the Solr enterprise search platform,and search engine interface on mobile music application are simulated via a Web application.Based on user rating or operating data,considering the effect of active users and popular items in the collaborative filtering algorithm,the different phases of the collaborative filtering algorithm are considered separately,aiming at relieving the Matthew effect and popularity bias phenomenon.Based on the analysis of the user behavior log in the mobile music application,this paper proposes a method for music recommendation,that is,a hierarchical recommendation method based on user interest.With measuring different users' interest preferences,we can apply different recommendation strategies to form a hierarchical recommended result to improve the accuracy and personalization of recommendation for users.