| Nowadays,the rapid development of the Internet makes the public opinion spread faster and more widely on numerous social networking platforms.The sentimental analysis of social network text can effectively reflect the development and change of public opinion,and the prediction and study of public opinion development can better support decision-making and effective processing.Therefore,this thesis has important research significance and application value.At present,the sentimental analysis of text is mainly focused on single text,the sentimental polarity of text is judged by mining the sentimental features of the text.There is a lack of integrated analysis of the internal information of multiple features such as sentimental features and textual features.Moreover,most of the existing methods focus on the features of the text,and the research on fusion of the time sequence of sentimental features and polarity information is deficient.Therefore,this thesis crawls the public opinion text information in the field of P2P Internet loan enterprises from the social media network platform on the Internet,including microblog and website of home of Internet loan.On the basis of time series analysis,this thesis discusses the fusion mode of emotional features and text features,and propose and realize a multi-feature model using the self-attention mechanism,make use of the influence of time dimension and the importance of different feature directions.This thesis realizes the prediction and analysis of the sentimental situation of the public opinion text information of the social network of P2P Internet loan enterprises.Through the experiment of the real data,we show that the value of F1 of fusion features in microblog data and Home of Internet Loan data is better than that of using sentimental features alone or textual features alone.The effectiveness of the system is verified.At the same time,in order to reflect the long-term trend of public opinion,this thesis constructs a multi time window public opinion trend prediction model based on the public opinion text of P2P network loan enterprise field,which can realize the prediction of public opinion trend for a longer time.The results show that it can effectively depict the phased trend of public opinion.Finally,based on above studies,a prototype system of public opinion situation research is designed and implemented. |
