The Social Network’s Nodes Influence Evaluate Based On Sentiment Analysis

Social Network Service is a platform based on friends’ relationship, where people can release, share and spread information. It contains abundant uses’ personal information and links between them. Opinion leaders in social network play an important role in filtration, diffusion and communication, their credibility and influence can not be ignored. How to identify them effectively is a hot spot in this research filed. Evaluating the influence of social network’s nodes reasonably, identifying the opinion leaders scientifically according to the result, it is significant to social management, business marketing and so on.Based on a review of relevant domestic and foreign literature, this article takes Autohome as research object, combines with social network analysis and sentiment analysis, proposes a new method by using three-stage to identify opinion leaders. First, considering the network structure, pre-ordering users influence by evaluating users activity, we get the target users list. The first stage references to the basic idea of PageRank, studies the characteristics of user’s behavior, chooses several representative coefficients, designs a mechanism of uneven distribution to UA Rank among uses based on multi-index comprehensive evaluation methods. Second, considering the social attribute, we identify the target users’ emotion tendency through interactive text. In the second stage, an effective and overall polarity lexicon is constructed,including four lexicons, about base lexicon, domain lexicon, modifier lexicon and network lexicon. We study polar strength by changing the relative position between adverbs of degree and negative words. Based on the polarity lexicon and operational rules, we calculate polarity phrases, then use polarity phrases as basic unit to calculate users’ emotion polarity value. Third, combining network structure and social attributes, we design the final evaluation model of node influence. We adjust the two weights for the first two stages to be suitable for different application scenarios and to meet the needs of different experiments.In order to support research, we take Autohome as a case study. It has been proved that the three-stage evaluation model is accuracy and efficiency. By this method, we get opinion leaders, who are at the core of the network structure and really get the support and advocacy from other users.In a word, we combine network structure and social attributes to meet dual nature of social network, and propose innovation in every stage. By the three-stage evaluation model, we improve the efficiency and accuracy, overcome difficulties that sentiment analysis does not apply to large-scale social network analysis, and make the result better meet the objective reality.