The Research Of Feature Selection Method And Sentiment Analysis Based On Microblog

Microblog is nowadays becoming the most popular Internet application. It has formed a highly interactive platform for information dissemination on a global scale. Its simple operation, freedom of expression, quick update features quickly developed into a multibillion contains a huge community of users. Globally, there are more than700million micro-blog updated daily. In China, there are a few users up to500million microblog. More and more users are keen to express their views and opinions on the microblog, to express their views, attitudes and emotions. The huge amount of information inevitably brings in information processing and information filtering problems. At the same time, changing the topic of microblog appears to be fragmented and no rules, in fact implies a huge commercial value, but also has important implications for public opinion monitoring and other government departments to work.Based on that, this thesis focuses on the feature selection method and emotional analysis of both studies. For the study of micro-blog feature selection method, the text was first proposed based on improved information gain feature selection algorithm for improving the precision of feature selection. Later the thesis proposed a microblog based on the dynamic characteristics of the data stream, select Update method is well suited for feature of big data. For the study of microblog emotion analysis, the thesis first constructed emotion dictionary. Then we design a propensity analysis method based on microblog emotional syntactic dependency approach.Experiments on SinaWeibo corpus show that the feature selection, the proposed method than the traditional information gain in precision and recall rates were increased by4%and3%.Meanwhile, the dynamic signature updates designed more applicable to the case microblog data stream. For the emotional analysis, sentiment analysis method based on syntactic dependencies is applicable to microblog sentiment analysis judgment. Positive tendencies microblog precision is63%and the negative tendencies microblog precision is78%.Experiments show that the method is suitable to solve the problem of Chinese micro-blog sentiment analysis.