Quantification Research Of User's Influence In Online Social Network

As a kind of ?instant information sharing platform?, Micro-blog has showed some outstanding advantages comparing to the traditional media, and has brought great convenience to social life. Evaluating the influence of Micro-blog users reasonably and effectively can produce both high theoretical value and huge economic benefits. Therefore, it is significant to study the influence of users on the Micro-blog platform.Because the traditional user influence evaluation method is lack of the important micro blog text dimension, and neglects the problem of industry domain, this thesis presents an evaluation method based on micro blog text in different domains, which is mainly composed by following:Firstly, based on the linear weighted quantitative model, the thesis presents a quantitative model for the influence of the micro blog text, which is a comprehensive assessment of the number of retweets, the number of comments, and the distribution of the information.Second, for the traditional research ignored the problem of domain, this thesis propses a CDo L&K(Algorithm of Classifying Micro-blog Domain based on LDA and KNN) classification algorithm, based on LDA(Latent Dirichlet Allocation) and KNN(k-Nearest Neighbor). The algorithm uses Gibbs sampling the LDA model in the data set, and then classified the micro blog texts with the KNN algorithm. At last the number of texts and weight in different fields of the users can be calculated.Third, combined with the above two points, this thesis proposes an ACIU(Algorithm of Calculating the Influence of Micro-blog User) algorithm based on the modified CIUA(Calculating the Influence of Mico-blog User Algorithm) algorithm. This algorithm is more accurate to measure the characteristics of the micro blog users and the characteristics of the micro blog text. And evaluating the user influence in the Micro-blog platform is more effective and accurate.Finally, the method to obtain the various types of data is introduced, and comparison tests between the traditional and the proposed method are done. Results show that the method of evaluation in this thesis is more practical and reasonable.