Research on Minimizing Misinformation Attention by Nodes Blocking Strategy in Online Social Networks
Information in online social networks influences people's views and opinions.Misinformation mixed in it is bound to mislead people's judgment and decision-making,and is particularly likely to cause people to panic,dissatisfaction and other emotions.The more people pay attention to misinformation,the more likely they are to be misled by it and engage in irrational or even aggressive behavior.When dealing with misinformation,we should try our best to avoid situations where misinformation leads people to engage in extreme behavior,that is,to reduce users'attention to misinformation as much as possible.However,minimizing user attention to misinformation is different from minimizing the number of users affected by misinformation,and it is a new problem that existing methods or technologies still face some difficulties in tackling.Therefore,we propose the misinformation attention minimization problem by nodes blocking strategy,where the goal is to minimize the total attention of users to misinformation when they are activated by misinformation.Firstly,this paper takes the classic information propagation model-the independent cascade model as the basis,and considers the user's attention to misinformation to construct an attention cascade model for the spread of misinformation.The Coulomb's theorem,which describes the interaction force between stationary point charges,is leveraged to characterize the evolution of users'attention to misinformation during the spread of misinformation.Secondly,the NP-hard of the problem of minimizing the misinformation attention by the node blocking strategy in social networks is demonstrated.The non-negative and non-monotonicity,non-submodularity and non-supermodularity of the objective set function of the problem of minimizing the misinformation attention by the node blocking strategy,as well as the#P-hardness of the computation are also verified.Thirdly,we introduce a parameter-attention reduction value,which describes the reduction in the attention of activated users to misinformation after blocking the user set.Based on this,the problem of minimizing the misinformation attention by the node blocking strategy is equivalently converted into the problem of maximizing the attention reduction value by the node blocking strategy,and a(ε,δ)-approximate Monte Carlo simulation method for estimating the attention reduction value of misinformation is given.Then,we leverage the supermodularity ratio parameter of the set function to explore the approximate properties between the Lovász extension function of set function and it concave closed function,and propose an approximate projected subgradient algorithm.Finally,the effectiveness of the proposed algorithm and model is verified on three real data sets:YouTube,Facebook and Digg.Experimental results show that under different experimental settings,the approximate projected subgradient algorithm is consistently better than existing heuristic algorithms in reducing users'attention to misinformation(at least 11.35%),and is at least 19.05%better than the baseline algorithms in suppressing the spread of misinformation.In addition,our numerical findings that the initial attention vector of users in social networks affects the decreasing value of users'attention to misinformation in social networks,but has a certain homogeneity,and users'attention to misinformation affects the efficiency and effectiveness of misinformation governance.
万方数据
