Guba Comments Emotion Classification Based On String Kernel

The traditional finance theory is put forward on the assumption that the shareholders are rational,but this theory is only suitable for more mature stock market.That is to say,shareholders can rationally judge the stock market and can correctly interpret the market information.Obviously,for the Chinese stock market,the demands are harsh.Chinese stock market is still at the growing stage,and there will be soaring plummeting,so behavioral finance experts carried out some related research.The result of research shows that the investment behavior of irrational shareholders can directly affect the final stock movements and whether the investor sentiment is high is also closely related to the stock market.Therefore,analyzing the emotional changes of shareholders is very important to analyze the stock fluctuation.In the Internet age,most shareholders can communicate and obtain information online.Guba is a stock exchange platform and the participants are mainly shareholders.Shareholders can share investment experience and express their opinions through guba.At the same time,it also provides a platform for analyzing the current emotional tendency of shareholders.This paper studies the emotional tendency of guba(oriental wealth network shares)comment text.First crawl the comment text through Scrapy crawler technology,filter the text and vocabulary.Then,replace the text with synonyms based on LSA and PageRank synonym recognition algorithm.Finally,use the proposed MSK string kernel to check the comment for emotional classification based on support vector machine(SVM).Through the experiment,the comment classification effect of MSK string kernel is better than traditional string kernel function and common kernel function.The main research contents of this paper:1.Synonym recognition algorithm based on LSA and PageRank is proposed.This algorithm combines LSA and PageRank algorithms.It retains the advantages of semantic mining in LSA and integrates syntactic structure information through PageRank algorithm.Finally,through analyzing the experimental results,it is found that combining the two algorithms can effectively improve the efficiency of synonym recognition.2.The MSK string kernel function is proposed.Firstly,the algorithm of subsequence weight based on string is proposed.Subsequence weight is composed of subsequence compactness and subsequence importance.When using string kernel functions to classify text,by improving the kernel value calculation method of traditional string kernel function based on subsequence weight,MSK string kernel function is obtained.Experimental results show that the text classification effect of MSK kernel function is better than that of traditional string kernel function and common kernel function.