Research On The Analysis Of Weibo Comments Tendency Baesd On Machine Learning Methods

This paper focuses on research object of orientation analysis of Weibo comment. The main research content is Xinlang microblogging reviews for the study and the emotional tendency. The research refines several features of the study and through improved integration of machine learning methods to enhance the accuracy of classification.Sentiment analysis has been widely used in monitoring public opinion and commodity inspection. Based on this, the paper proposes an idea that divides the comment into three types, which contains spam comments, subjective or objective comments, and appraise or critical comments. To select the appropriate mode for analysing different types of comments different types of comments.Firstly, the paper clean up the data of comment that peels off the spam comments and objective comments. The paper uses effective integration of several features, Naive Bayes and the method of dividing threshold to determine spam and objective comments. This method greatly reduces the noise of the comment.Secondly, contrary to the comments of appraise or critical, by comparing several feature extraction methods and improve emotional words based on its selection method and calculated weights method to constitute a new comment vector space. By combing machine-learning methods and voting method to merge traditional machine learning methods. This method can achieve a better analysis effect. This paper implements the emotion words based on its selection method and calculated weights method, and use Ada Boost, Random Subspace and fusion classifion to enhance traditional machine learning methods, which can improve the accuracy of the analysis of comments.Finally, through the performance evaluation methods describe the method for the analysis of public opinion has a good effect. The paper expands emotional shift and emotional carrier for the comments, juding similarities and differences of emotional carriers of appraise and critical dataset, which can play a role in early warning public opinion.