The Weibo Account Detection Based On HMC+ Weighted Bayesian Neural Network

With the continuous development of network technology,people's social approach has gradually expanded from real life to virtual social platforms,and Sina Weibo has become the preferred social platform for most people.People can get a lot of information from Weibo through Weibo account,but there are many zombie accounts in these Weibo accounts,which often publish a lot of fake information,spam ads and phishing websites.Traditional account detection methods have been unable to detect abnormal accounts quickly and efficiently.Therefore,the existing account detection algorithm should be improved.In this thesis,the focus of research is on the research of microblog account detection algorithm.Through the research and improvement of account detection algorithm,the correct rate and running speed of account detection algorithm are improved,and the computational overhead is reduced.This thesis proposes a combined model of HMC+ feature weighted Bayesian neural network to detect microblog abnormal accounts.The main research contents of this topic include:(1)In order to reduce the impact of noise interference on account detection,this thesis uses feature-weighted Bayesian algorithm to give different weights to each behavior feature,so as to reduce the impact of noise on Weibo account detection.(2)As the number of accounts increases,the structure of the account detection algorithm needs to be optimized and improved.This thesis improves the structure of the account detection algorithm by feature-weighted Bayesian neural network.(3)By combining the Hamiltonian dynamic system and the Markov-linked Monte Carlo method to form a hybrid Monte Carlo(HMC),the convergence speed of the Markov chain Monte Carlo algorithm is improved,and the speed of account detection is further improved.