Theoretical Model Investigation On Massive Music User Behavior

With the development of information technology, information interaction through the Internet becomes more popular. Users generate huge amounts of data during the interactive process, which contain a lot of valuable user information. Large databases are used to collect and store the fast growing mass of data. Therefore, it is extremely important to do data processing and data mining by taking suitable computing techniques.The Hadoop technology, which is based on distributed computing principles, is capable to provide data access and processing on high throughput data. This thesis described the use of Hadoop technology to process the mass data of music user behavior. According to the characteristics of music users and the important indicators that music enterprises focus on their website users, six dimensions as user behavior attributes were selected to be analyzed:full site accesses, top list page views, event page visits, the number of songs played, and the number of songs downloaded. As the data source, the user access log recorded by web servers were selected, as well as the behavior log tagged by JavaScript. By using Hadoop technology to code MapReduce parallel computing program, the mass data of users are being preprocessed and then computed. As a result, a multidimensional model of music user behavior from mass data was established.This thesis used K-means clustering method, which is based on partition, and the K-medoids clustering algorithm, to analyze the multidimensional model on music user behavior. The users were clustered into6classes by considering the SSE of objective function and the clustering effects and other factors. Through the comparison between different types user behavior, the paper proposed a more targeted approach to guide different types of users, in order to improve the user visits and songs playback amount, increase user stickiness and cultivate more loyal users. Moreover, it will help the music companies to grasp and study the user’s overall changes in a timely manner, to find out the characteristics and laws, to provide more targeted personalized services for different types of music users, and eventually for them to increase profits and market share.