Research And Implementation On Feature Selection And Emotional Analysis In Public Opionion Analysis System

Network public opinion refers to different positions and comments which are held by the people on the Internet for social focus and hotspot issues.Nowdays,more and more people begin to use the Internet to express their emotions and views.And network has gradually developed into a collection and distribution center of the social public opinion.Due to the relatively loose network environment and the lack of corresponding regulation measures,some people tend to regard the network as a place to vent their dissatisfaction,venting their discontent wantonly,and guide the wrong direction of public opinion.Besides,our country is in the stage of social transformation,and full of all kinds of contradictions,some social management departments often tend to hide and avoid public opinions,which leads to concentration of social contradictions broke out.Nowadays,under the background of the era of big data,distinguishing the public opinion information on the Internet through the traditional artificial way is not only inefficient but also not realistic.Therefore,it needs the network public opinion analysis system to obtain and analyse the mass information on the internet,and warn the misleading opinion early in time,witch is very important to the development of the social various aspects of China.The author of this thesis studied some techniques involved in the network public opinion analysis system.The main work and innovations are as follows:1)The author of this thesis introduces the realization principle of each module of public opinion analysis system from the theoretical aspect.Besides,the feature selection algorithm and text sentiment analysis technology are studied as two basic research points.2)It analyzes the shortcomings of four traditional features selection algorithms,which are the document frequency,information gain,mutual information,and chi-square.And on this basis,it also puts forward a kind of the feature selection algorithm,which based on the average degree interclass of feature and the distinction degree between class of feature.Finally,the experimental scheme is designed and the efficiency and rationality of the improved algorithm are proved.3)Based on the research of text emotion tendency technology,it designs a method of text emotion tendency analysis based on semantic.The emotional words in the subjective sentence determine the direction of the sentence emotion,and some other modifiers determine the degree of sentence emotion.The method is based on subjective sentences-emotional words.All the subjective sentences in the document are selected,and the emotional tendencies of each subjective sentence are scored respectively,and then the average value is taken as the emotional score of the text.Through the experiment,it is proved that the text emotion analysis method designed in this thesis has a good effect on the judgment of positive emotion and negative emotion.4)Using JAVA language implements a public opinion analysis system based on B/S framework.