| False network public opinion not only confuse the public viewing,but also easily lead to bad social mentality,thus intensify social contradictions and affect social harmony and stability.In addition to strengthening the legal and institutional construction in the field of public opinion,it is necessary to use technical means to monitor and intervene it for the false network of public opinion.The detection and Inhibition algorithms of the false network public opinion based on web forum are studied systemically and deeply in this paper.The main contributions are as follows:1.Based on the node influence,an opinion evolution and public opinion formation model of discrete time series was proposed.The model is based on French-DeGroot model and fully considers the continuity of the change of nodes’ opinions in the network,as well as the connection strength between nodes and their timevarying characteristics.Simulation experimental results show that the model is in complete accord with the opinion evolution and public opinion formation of the online network.At the same time,we can get the opinion profile and other statistical properties from the model at all time.2.Based on the noise filtering and topic clustering,a fast method for bursty topics detection and tracking was proposed.Its basic content can be summed up as follows: Firstly,the preprocessed web forums data are generated for the topic candidate set.Then multiple filtering is used for the topic candidate by the predefined hotness and reply acceleration indicators.Finally,bursty topics are detected by using the topic clustering algorithm.The results demonstrate that the proposed method can get higher precision,recall and F1 in bursty topics detection.Additionally,it can effectively track the bursty topics,which makes up for the shortcomings of traditional static methods.3.A method,which transforms the detection issue of false public opinion to the classification issue of the quad corresponding to the bursty topics,was proposed.Through analysis the characteristics of water army driven burst topics in their latencies,a bursty topic is defined as a quad containing the replies index,new registration ID index,simple replies index,ID discrete index.Since the differences between normal quad and abnormal quad on their characteristic elements,a novel bursty topic classification algorithm based on SVM active learning was proposed.The algorithm takes the initiative to choose bursty topic which it thinks is with the maximum uncertainty of the current classification and trains the sample to further design the new classification function,so that we can use as little as possible thenumber of labeled samples to achieve the highest possible classification accuracy.The experimental results show that the detection accuracy of the proposed algorithm is significantly better than the composite indicator method.Additionally,in the case of less labeled training data,the algorithm for detection efficiency and detection accuracy is also superior to traditional SVM algorithm.4.Based on the heterogeneity of web forum user behaviors and SEIR model,a public opinion spreading model for the web forum was proposed.The model takes into account the degree of intimacy and trust between web forum users,the state of web forum users and the transition relation between states.Simulation and numerical analysis shows that the model can not only characterize the public spreading and evolution of the real web forums,but also the introduction of trust mechanism is capable of effectively reducing the public opinion influence,the velocity of the public opinion spreading and the public opinion size.5.Based on the high-influence user immunization algorithm,a public opinion inhibition algorithm for web forum was proposed.Based on the above simulation and numerical analysis conclusion,the algorithm first constantly seeks the most influential nodes from the selected node and its adjacent nodes through reiteration method,then it immunes them one by one,that disconnects them from other nodes,so as to achieve the purpose of inhibiting the public opinion spreading. |