Evaluation Of User Influence In Social Media

In recent years, the development of Social Media on the internet has attracted a lot. Now, social media has been a Web2.0business which has the widest users, the greatest impact on dissemination of information and the highest commercial value. And more and more attention has been paid on social marketing based on social media. In this context, we propose to take field of users into the model of influence evaluation.The objectives of this research include gaining users’behavior patterns in social media; building the standard of field classification; establishing an acceptalbe model of text categorization; establishing the Model of Field Influence Evaluation (MFIE) which reflects the real influence of users in social media.This thesis includes literature investigation, user study, on-line data analysis, establishing the model of field influence and case study. In the first phase, the relevant research and websits on influence are gained by collecting and analyzing the related literature. In the second phase, questionnaire survey method is used to prove hypothesis of behavior patterns created in interviews. And the result is to be used as the basis of establishing the standard of field classification in on-line data analysis. In the third phase, content anyalysis is used firstly to anlyze and code the data deeply.Combining with the exsit standard of classification, the standard of field classification is built. According to the standard of field classification, a corpus is set up for Machine Learning. To build an acceptable model of text categorization, we pretreat the data, improve the standard of classification and enlarge the sample of corpus. In the phase of establishing MFIE, we divide related data into information and relationship data firstly. Then based on depth of field, MFIE is established. In the phase of case study, the field influence of different users in one field and one user in different fields are compared.In the research, we come up to the concept of Field Influence and indentify which field the users are in by the method of Machine Learning. This thesis generates new value on the research of opinion leader.