A Micro Blog Friend Recommendation Algorithm Based On User Tendency

With the social network being expanded in numbers, the numbers of users andthe volume of information are exploding, and the development of Micro blog is agood example. However, with the exploding data, a user is difficult to find a usefultarget among huge data. So all Micro blog web site have their recommendationsystems. The follow and followed friendship is one of the main channels to getinformation on Micro blog. However, it’s a very difficult problem to find a properuser to be followed among billions of users. Among mass of users, it’s hard to find adesirable friend without a good recommendation. Therefore, recommendingfriendship to build connections among users is one of the fundamental tasks for socialnetwork systems.In this paper, we are focused on the friend recommendation algorithms based onmicro blog. Firstly, the paper summarizes the existed algorithms used on friendrecommendation, and discovered that most of friend recommendation systems need toevaluate the users’ similarity in order to generate top K candidate recommendations.Secondly, we proposed the concept of similarity model which is used on friendrecommendation systems and explained the faultiness of traditional recommendationalgorithms. At the same time, we analyzed the features of users’ relationship on microblog. Thirdly, we propose the similarity model based on the users’ tendency. Themodel can fit the users’ behavior on micro blog which can down to contact andinterest behaviors. We implement a new recommendation algorithm based on thatmodel. Lastly, the paper imports users’ trust mechanism into our algorithm, the usernode can propagate its similarity to others nodes. After propagation, the algorithmgets the final recommendation results.Experimental results show that the users’ tendency similarity model can giveexpression to users’predisposition on micro blog. Each user has their alike and dislikeon other users. The algorithm based on the users’ tendency similarity model achieveshigher performance than the traditional recommendation algorithms. Applied the users’trust mechanism, the algorithm achieves a higher performance.