Selection Algorithm Based On Mixed Emotional Features Pawlak Attribute Importance

With the rapid development of Internet people’s lifestyle has changeddramatically. Static Web pages are no longer content the needs of people informationtechnology brought us to a completely user-centered year. However, with theimprovement of life quality, life rhythm is speed up, more people tend to sendmessage in the micro-blogging platform,which can contain Casual format andreal-time interactive information.This quick and easy means of communication notonly conforms to the development of contemporary Internet,but also gradually affectour lives.sentiment analysis research on micro-blog content is becoming a hot researchtopic, because of the convenient and widespread. In the study of sentimentanalysis, there are still some problems, field dependency, semantic comprehension,feature extraction, restricting the number of issues with the recall ratecharacteristics.This paper research sentiment analysis on micro-blog, and introducesthe main method of sentiment analysis. Using supervised machine learning methods,found that the traditional method is always single and not good at feature selectionfor micro-blog mining. So propose mixed feature selection method based on roughset theory IGRough and CHIRough. We support experimental results showed themixed feature selection method have high efficiency, greatly improve the recall ratetext and F-value obviously. We also found stochastic gradient descent algorithm isvery useful method for sentiment classification. Following we combine variousclassification algorithms and feature selection algorithm to find best model.At last paper give an example using sentiment analysis method to identify micro-blogsentiment for public opinion analysis.