Research On Constructing Emotional Incremental Model Based On Dual Features Coevolution

With the rapid development and popularization of Internet technology,it has been integrated into people's daily work and everyday life.It has largely replaced the traditional information media which represented by television and newsthesiss and has become the mainstream media of information in the society.In recent years,due to the spread of social networks both at home and abroad,major social events have multiplied.They have brought tremendous pressure on the country's public security and social stability.Therefore,analyzing the user's emotional state in social networks deeply,revealing the regularity and essential characteristics of the communication between users' emotions,which can provide opportunities for managing,using and correctly grasping the orientation of public opinions.At the same time,it has extensive and realistic significance for purifying the network environment.According to the traditional network emotion analysis,only the emotional characteristics of network nodes are considered,while the social characteristics of network nodes are ignored.This will lead the influence of the surrounding environment on the emotion of the user is neglected and in many cases,the user's emotional mutation is closely related to the user's influence.Therefore,the prediction accuracy of the user's emotion will be greater fluctuations.This thesis proposes an emotional incremental model based on dual features coevolution,which will affect the emotion of network nodes in social relations and historical emotions.Because emotion is evolving,its evolution process is a relatively slow process.It can achieve the goal in an incremental way without analyzing all the historical data to obtain the emotional evolution model.Therefore,this thesis proposes the calculation of affective decay of network nodes and the setting of the parameters of affective penalty and gain of network nodes based on social characteristics,which affect the affective effects of nodes and the social relations on their own emotion.Then,the co-evolution of the two aspects of influence to build a new emotional model.This will not only improve the emotional prediction accuracy but also greatly improve the stability of computation.In this thesis,a gradient descent algorithm based on probability factor graph is used to train the model on the data captured by some social networks such as Weibo and Renren.The trained model predicts the remaining data.Through the chaos matrix established by the experimental results,it is found that the emotional incremental model based on dual features coevolution proposed in this thesis is better than the traditional SVM classification method in terms of accuracy,recall and other measures of the model,and the data is less.Therefore,the emotional incremental model based on dual features coevolution is more accurate than the traditional emotional analysis model,the stability is higher and the processing result is more credible.