User Recommendation Algorithms Based On Micro-blog Platform

As a network platform based on user interest information sharing communicating and accessing,Weibo(Micro Blog) rapidly become a mainstream form of social networking by its unique advantage of the spread.Through the microblogging platform, users not only can get the information they are interested in, but also can maintain friendships, make new friends, expand their social circle by using the various functions provided by the platform at the same time. But with the development of Micro Blog, a sharp increase in the number of users, so as the amount of information contained in the microblogging platform. Faced with such complex data, how to extracted useful information from these repetitive overload of information and use it in Weibo user recommendation is a serious problem need to be solved.Although the technology of user recommendation on social networks has been penetrate deeply researched by scholars, which have achieved great success. But the results are not perfect, and research of recommendation on microblogging platform also can be improved. So based on the characteristics of the user social information from the microblogging platform, this paper study the user similarity calculation method and propose improvements. In this paper, the main contents are as follows:1. This paper analyzes the current status of research on microblogging platform and user recommendation on it. And in this way on the basis of the characteristics of users interest connected on the microblogging platform, we will analyze the feasibility and necessity by using Weibo user social information to recommend. Then details a recommendation algorithm based on two-stage model of user recommendation using user social information.And point out that it`s not accurate enough for the result in the calculation of the user similarity because of the undifferentiated similarity calculation. And based on that this paper proposes a method to calculate the cosine similarity join an improvement factor of the popularity of checks and balances.The method of calculating the similarity based on user interest information to distinguish the different effects from popular star user to the ordinary user. In the measure on the basis of the original cosine similarity, we adding popularity balance factor to reduce the different effects on the results of user similarity calculation based on the information concern. And then presents an improved basis for this recommendation based on the user's attention to the information the user algorithm in two stages.2. This algorithm will be presented by the use of user recommendation experiments on Sina Weibo user data set while using the original algorithm and "Friend-of- Friend" algorithm as the reference algorithm. And compare the performance of the algorithm recommended by both the overhead and time. In order to analysis the experimental results of the recommendation algorithm`s performance, we use the accuracy, recall and comprehensive recommendation index. Experiment results show that compared to the reference method, so that it improved the recommendation accuracy.