Social Network Data Mining Based On SAF Model

With the rapid development of network, the social network becomes more and more complex. People begin to put the data mining skill into the social network to search deeper-level, more valuable knowledge. Building a model which can make full use of the information in the social network is a great significance in improving the effectiveness of the social network data mining. This article’s study is based on the e-mail network which is one type of the social network, it includes abundant information:communication relationship, communication manner, communication frequency and so on.AF model is an activation force measure model proposed by Pro.GuoJun. This model has achieved some excellent results in word network and protein-protein interaction (PPI) network researching. This article combines the characteristics of AF model and the e-mail network, redefines and resets some variable of it, then proposes SAF model in the e-mail network.Firstly, this article studies the community-discovery problem in the social network, applies the two most core measures (activation force and similarity degree) into community-discovery. After grasping the importance and difficulty of community found, this article achieves a total of three algorithms:classic GN algorithm, GN algorithm based on activation force, community-found algorithm based on similarity degree. The results shows that it can achieve better effect to delete some parts of the edge nodes through activation force before run GN algorithm, the similarity degree method structures the nodes to be a tree first and then generates sub-community by way of pruning.Secondly, this article studies link prediction problem:searches the most similar nodes through similarity degree and then searches the closest nodes through activation force. Combining with these two methods, the algorithm in this article has been built. It proves that the algorithm in this article is effective by seeing two indicators:accuracy and recall rate.Finally, this article studies identifying problem of the core person in the community, establishes an "attention degree" model, researches from two point of views:communication frequency and activation force. Analyzing the results, it can prove that it is the most effective to digging core person by activation force.