Detection Method Of Malicious User Groups Of Online Financial Forum And Its Application

Recently years,the rapid development of the Internet promotes information and network communication technology.The social life with lots of information contains the value of the data on the network.And social networking media which exists vast users,has been widely used for auxiliary decision-making by the organizations and individuals.There are huge users and potential business opportunities in online financial forum,and it makes false opinions and spam manufactured and widely spread possible.And the source of this kind of harm comes from malicious user groups.In view of the above problem,we use web information extraction technology,storage network user behavior data,sentiment analysis,network modeling,overlapping community detection and evaluation technologies to collect online financial forum user behavior data,model the user network with the relationship between users,detect the communities from the network of the user,testing the malicious user groups and evaluation the results.The main work of this paper is as follows:1.Research the structure of the online financial forum website page,analysis forum user behavior,use the web information extraction technology,match and collect behavior data of the forum users,and store the data in the local relational database in My SQL.2.Based on machine learning,split word of the training set,select feature,choose the appropriate emotion classifiers,classify user comments on the content of emotional prediction,based on the predicted classification results,build network user behavior model,and describe the user similar emotional network related global statistical characteristics,draw that the similar emotional network meets the "small world" properties,and also has the scale-free characteristic.3.Put forward an overlapping community detection algorithm which is based on the topology and the similarity of node attribute,use this algorithm to test the online financial forum user network and three social network data sets of Stanford University for overlapping community detection,and compare with common community detection algorithm.4.Put forward the corresponding external index of the community detection and combine these external indexes to identify malicious user groups in the stock forum,and analysis them with case studies.