Weibo Malicious User Detection Method Based On Behavior Feature Analysis

With the booming of social networks,weibo provides a fast and reliable online social communication channel for domestic social network users with its large frequency of information generation,fast propagation speed and wide range of sharing.The emergence of Weibo has been welcomed by more and more people.However,a large amount of malicious information and malicious users appear on the Sina Weibo network,affecting the normal users 'experience of using the products and also causing serious problems to the normal users' lives.This thesis obtains the malicious user and normal user data through multiple channels.The Principal Component Analysis(PCA)algorithm is adopted to extract the behavioral features.The weight of each dimension feature is sorted.It also makes microblogging malicious users and normal users better distinguished.At the same time,based on improving the classification algorithm,this thesis proposes the classification model PCA-Random Forest.Through the establishment of a malicious user identification system,the malicious user identification of microblogging is completed.(1)To study the extraction and analysis of malicious user behavior characteristics,this thesis refers to use the PCA algorithm to perform dimensionality reduction analysis on the behavioral characteristics,and finally extracts some high-resolution main component features: the six-dimensional principal component features of PCA1-PCA6.Three-dimensional new features are fitted through the indicator function using the principal component features.(2)To establish the PCA-Random Forest classification model,this thesis introduces stochastic forest classification algorithm to identify malicious users of Weibo and improves the efficiency of the overall classification model.Finally,the best feature model selected by PCA algorithm is merged with random forest algorithm to construct the classification model: PCA-Random Forest.(3)To build a prototype system,this thesis designs a prototype system to realize the malicious user identification of microblogging.The system tests the classification system,and compares the basic functions of the system with the feature dimension and the algorithm dimension.