Research On Dynamic Portrait,Prediction And Tracking Of Hotspot Topics Of National Security Threats On Social Networks

Social network is a kind of tool for information sharing and online communication based on user’s social relations.With the development of mobile Internet in recent years,users can be more easily to build up connections and more frequently to use social network.Social network itself has four characteristics:agility,diffusivity,equality and Spontaneity.Because of these characteristics,social network has already owned billions of users in the short decades since it’s born.However,with the increasing number of users,the power which is used to control the Internet information has been gradually distributed to everyone rather than the traditional media which is represented by a few official media.That leads to a problem of information unreliability.If the information are those that users generally don’t care about,they won’t be spread widely.However,Once national security is involved,the unreliable information of national security will affect and impact social stability.On the other hand,both domestic and foreign there are some suspicious forces using social networks to release rumors and try to incite riots.It can be analyzed from these real cases that timely and effective guidance of public opinion is an important way to avoid public opinion out of control.Fast and accurate portrait of the hot topics of national security threat,tracking and predicting the development of public opinion enthusiasm can effectively help managers to track the direction and status of public opinion,so as to timely guide it.The main work of this paper includes the following four aspects:Deep semantic feature learning of national security threat hot topics in social networks.Research and solve the problem of sparsity and ambiguity of short text in social network,and expand the text reasonably;This paper researches the convolutional neural network models such as Alex-net,VGG-net and the attention mechanism.Through the modification of the existing network model structure,it achieves the deep semantic feature learning of the image,and trains the national security related image classifier of social network,which is the basis of the unified semantic expression of the social network national security threat hot topic in the next step.Semantic unified expression and dynamic portrait of social network national security threat hot topics.This paper researches on semantic space unification of social network multimedia big data based on attention mechanism feedback.The attention mechanism is used to strengthen the semantic attention in the image and suppress the noise in the image.It makes the semantic description extracted from the image more accurate;This paper also proposes an algorithm to automatically extract the candidate keywords of topics in different time windows,and extracts the topic keywords based on the topic model,and finally obtains the hot topic portraits of social network national security threats in different time windows.Prediction and track of social network national security threat hot topics.This paper researches the prediction and track methods of social network national security threat hot topics.According to the statistics of public opinion’s heat of different topics in different time windows,the public opinion’s heat vector and public opinion’s heat change vector of the same topic at different times are constructed by the time sequence.According to the change of the public opinion’s heat with time,it is divided into four states to track the development and change of public opinion’s heat,and the neural network process is used to train the mapping relationship between interaction data and public opinion’s heat of social network.In this way to achieve the prediction of public opinion’s heat.The achievement of the dynamic portrait,prediction and track system of social network national security threat hot topics.Combined with the above three aspects of research content,developing the social network national security threat hot topic dynamic portrait and prediction system.Using this system to verify the accuracy and real-time performance of the algorithm proposed by this paper and provide visualization services in the topic portrait and heat prediction results.This paper which focuses on the data of social network national security hot topics achieves the dynamic portrait of hot topics,tracking and predicting heat of hot topics’public opinion.