Research On Key Technologies For Analyzing Influential Communities And People In Social Network

The emergence and popularization of social networks resulted into a large amount of data generated through human interactions and links formation.Such bulk of data has attracted researchers'attention to extract interesting patterns and trends that could be incorporated as utilitarian knowledge for decision making.There are two kinds of factors for social networks such as community and people.Recently,community analysis has witnessed enormous research advancements in the field of social networks.Moreover,extraction of influential people still needs to be properly explored.Many models and algorithms have been put forward in order to extract the influential entities from respective domains.However,a little effort is made to shed light on prediction and ranking the communities while considering their levels that in real terms are determined by the quality of their published contents.Therefore,this thesis focuses on the social network analysis from the aspects of communities and people.The detail is listed as follows:1.Presenting a PageRank based ComRank algorithm for identification of influential communities in social networks.Specifically,ComRank measures the influence of communities and ranks them by incorporating joint weight that is composed of internal and external influence of communities.More precisely,the modification is actually made to evaluate the influence of the communities while considering the joint weight based transition matrix.Thus,ranking of the most influential communities is made by the proposed algorithm.Experiment results show significant improvements by presented algorithm in community ranking due to the inclusion of proposed weighting feature.The comparative analysis presents that ComRank extracts influential communities in terms of level while PageRank foretold some low level communities at high ranks.2.Proposing a TunkRank based IPRank algorithm for identification of influential people in social networks.First,an innovative interaction strength metric is introduced.Second,proposed algorithm considers the communications from perspectives of followers and followees in order to mine and rank the most influential people based on proposed interaction strength metric.Experimental results demonstrate that IPRank discovers high ranked people in terms of interaction strength while the prior algorithm placed some low influenced people at high rank.3.Effective features based on machine learning algorithms are proposed for prediction of rising venues in social networks.Specifically,five effective prediction features along with their mathematical formulations are put forward for extraction of rising venues.Second,for prediction purpose,four machine learning algorithms including Bayesian Network,Support Vector Machine,Multilayer Perceptron and Random Forest are employed.Extensive experiments are conducted to analyze the impact of each feature while examining the classification accuracy.Experimental results demonstrate that proposed features set is effective for rising venues prediction.The research findings of this thesis would be useful for decision making in viral marketing and advertisements or election campaigns.