The Discovery Of Opinion Leaders For A Particular Theme Based On Social Network Analysis

With the advent of the Internet and the rapid development of informationtechnology, the internet has been relative with people’s daily life. They can acquirevaluable information from internet, can communicate with others on-line withoutany meeting, can exchange feelings with their friends in different places, and theyalso can entertain in the internet. However, due to the quantity and heterogeneity ofthe network information bring about “information overload” and “knowledgeconfusing”, etc. So, they need opinion leader to give them some advice. At present,the research about the discovery of opinion leader still has some shortcomings,mainly in the following aspects: Without thinking opinion leaders are restricted byparticular fields. Without thinking the relationship between the tag and theunderlying theme, with the rapid increase of information, the difficulty and thedimensions of calculation are gradually increasing; which greatly improve thedifficulty for opinion leaders’ discovery. The present study is still based on smalland simple network analysis; the maneuverability is not good at large network.In order to solve the above problems, taking enterprises’ knowledge acquisitionin Web2.0as background, taking automobile forum as the research object, takingovercoming opinion leaders limited by fields and improving the quality of opinionleader as research purpose, this paper mainly from the following perspective:(1)The research of the latent topic discovery based on LDAThe article analyzed the latent semantic relations among the users, labels andresources’ comment, introduced the producing type of probability (LDA) model andmended it, through the methods of the latent topic discovery based on LDA to findthe latent topic from customer reviews, let labels map to a particular theme to showthe theme explicitly, the experiment results prove that this method can classifyusers’ comment by latent theme effectively.(2)The research about the opinion leader’s indicator systems based on thespecific topicAccording to the characteristics of opinion leader, using the social networkanalysis’ point centrality, betweenness centrality, closeness centrality, as well asusers’ activity, support and infectivity to construct opinion leader’s index systemfor a particular theme based on LDA latent theme analysis.(3)The research about the method of opinion leaders’ identification based onparticular themeAbove the indicator systems research, the article put forward a method torecognize opinion leaders, above the third chapter’s latent theme analysis, solvedthe problem of opinion leader limited by fields, and above the assignment in weight,the paper adopted maximum deviation method, overcame the defect of the previousresearch about weight assignment with strong subjective.(4)Particular case study for opinion leader recognition Based on the research of the latent theme analysis and the method of opinionleader recognition, taking information from automobile forum as data source,designed and developed opinion leader mining system based on SNA, studied thesystem’s technical framework and system structure, analyzed the functionalmodules of the system in detail, finally, realized the main function of the systemand gave some operational effect screenshot.