Research And Application Of Friend Recommendation System Based On The Analysis Of Weibo User Relationship

Nowadays, Social Networking Service(SNS), a product of the era of Web2.0, has become a part of our life. Compared with the traditional portal websites, SNS is dominated by users, users can browse the content on the internet, also can generate content. In a wide variety of social software, Weibo gets users' perference by its timeliness of news. Weibo's market share of Chinese social software is getting higher and higher, the frequency of people using Weibo is growing, especially with Sina Weibo outstanding. In Sina Weibo, individuals could follow the people they want to know, and the people who have similar interests with them. How to give users accurate recommendations about people they are interested in from massive users, is no doubt a research hotspot. However, at present, Sina Weibo friend recommendation attention focus on few aspects. This article studied the following directions and provide a more accurate friends recommendation algorithm for the users.First of all, the existing recommendation algorithms are introduced and studied, and several techniques for subsequent use are introduced: classification algorithm, clustering algorithm, Page Rank algorithm, and related concepts of ontology are introduced.Secondly, use the K-means algorithm to cluster the user tags that belong to the category of interests. Tags in the same category are similar. Users may be interested in all the tags that belong to one category. On the other hand, for data processing of the Weibo user's post status, after segmentation the text classification corpus of Sogou laboratory and stop words, use Word Net ontology library and self-build ontology library which is based on protégé to complete the feature word selection, after that use KNN classification algorithm to classify the user post status, and then combined the result with the user's tags and the interests of people that user followed, the user interests were judged, the judge have similar interests between recommended users and users to recommend.Again, the user within two degrees of friends, that is, between recommended users and users to recommend, at least there is a mutual follow between users or one-way follow to the users, through the interaction between the users and the relationship between users, an algorithm based on Page Rank algorithm is proposed to judge the influence of users around users to recommend. According to the user's sign information in Sina Weibo, through the geographical location and the type of location, determine the geographical location relationship between recommended users and users to recommend.Finally, integrated the above three kinds of relationships between the users, similar interests relationship, two degrees of friends relationship that can build connection, and similar geographical location relationship, in the design and implementation of a friend recommendation system based on the analysis of Weibo user relationship. The experimental data is provided by Sina Weibo open platform's API and Datatang, and the implementation of the system is completed by using java language. In the system, compared with several friend recommendation algorithms that current Weibo is using, Weibo friend recommendation algorithm based on user relationship analysis improves the accuracy of the recommendation.