Research On Topic And User Influence Prediction In Micro Network

With the interaction brought by web 2.0 technology, social network has been one of the most important social media. Micro-blog is one of the most popular platforms for information dissemination and sharing based on social network. The characteristic pattern of Micro-blog has triggered large amount of users' interest, which accelerates the frequency of the topic release and expand the scope of the topic transfer. That is a great challenge for academic research and application. Micro-bar is a post bar belonging to Micro-blog, which is of strong subject relevant. So Micro-bar has been studied as a topic in this paper. The topic population and user influence were mainly studied in this paper. The main jobs, conclusions and innovative points are listed below:(1) The topic population prediction model was built based on complex network. The topic characters related with topic population were abstracted based on correlation analysis and qualitative analysis. Then these characters were assigned to the model as input variables. Network weights were trained to predict the topic population. In the training algorithm, the self-adaptive rule for different crossover strategies was designed to ensure the quality of the mutation. The method can not only strengthen the monitoring of the mutation quality, but cover the shortage on convergence rate and individual disturbance for one crossover strategy. The main algorithm parameters were set through complexity theoretical derivation, which is of certain significance in practical application and theoretical research. In contrast with neural network, the structure was more flexible. And simulation results show that the model is more excellent in accuracy and stability with better adaptability.(2) Social influence prediction model was proposed based on user behavior, which combined with complex network and self-adaptive algorithm. To user influence, effective measures were defined and applied to user clustering, aiming to predict the user cluster to discover the influence. This method can not only improve the accuracy of prediction but also make it more convictive for influence definition.(3) In the user analysis, user influence, behavior characters and the role of users were analyzed by statistics, which has some value on personalized recommendation, business marketing and so on.