Research And Application On Personalized Rumor Refutation Algorithm Based On Graph Pattern

Intervention on rumors in social media to prevent the spread of rumors and reduce the negative impact,referred to as rumor intervention,is an urgent concern of major social media.In the existing researches,some studies believe that interfering with the influential nodes in the network can prevent the spread of rumors,but the granularity of this method is coarse.Other studies take machine learning to predict rumor spreaders,but this kind of method has poor interpretability.Therefore,this thesis proposes a personalized rumor refutation(PRR)algorithm framework based on graph pattern.PRR algorithm finds potential rumor spreaders based on users’ historical behavior,and realizes active,targeted and advanced rumor intervention for all users for different types of rumors,so as to reduce the negative impact of rumors.This thesis studies from three aspects: research framework,algorithm innovation and experimental verification.The main contents and contributions are as follows:Firstly,this thesis proposes a personalized rumor refutation algorithm framework.The framework mines rumor propagation patterns firstly based on reverse query technology,then identifies rumor audience based on regular tree,which includes pattern matching and candidate identification,and finally pushes rumor refutation information to rumor audience.Secondly,this thesis defines a rumor propagation pattern based on regular path,which is committed to capturing the commonness and characteristics of rumor spreaders,and proposes a reverse rumor propagation pattern mining algorithm calculated according to users’ historical behavior.Experiments on real data sets verify the effectiveness of the rumor propagation pattern and mining algorithm.Thirdly,this thesis proposes a batch pattern matching algorithm based on regular tree to obtain users who meet the rumor propagation pattern.In order to deal with the frequently updated large graph of social network,this thesis proposes an incremental pattern matching algorithm based on update pivot to improve the matching efficiency.Experiments verify the effectiveness of batch matching algorithm and incremental matching algorithm on real data sets.Fourth,this thesis take the COVID-19 data on Sina Weibo to analyze the algorithm.The experiments mining two kinds of classical rumor propagation patterns and calculating their rumor audiences,and prove the effectiveness of PRR algorithm by verifying the candidate nodes in two phases and visualizing the experimental results.To sum up,based on graph pattern theory,this thesis opens up new research ideas for rumor intervention on social platforms,puts forward a personalized rumor refutation algorithm framework,and employs real Sina Weibo data for experimental verification and empirical analysis to confirm the effectiveness of the algorithm and prove that it can provide reference for rumor intervention strategies on social media.