Empirical Study On The Relationship Between Emotional Word Frequency In Annual Reports And Performance Of Listed Banks Based On Text Mining

With the development of the Big Data,the amount of information people receive and the way they access information are constantly increasing.Various research reports,disclosure documents,even financial news can bring people the information they want,as an important information carrier,annual reports of listed companies are one of the important sources for both investors and business managers to obtain information about companies.The information in the annual report can be broadly divided into two categories,one is the financial data information composed of numbers,the other is the non-financial data information composed of words.The non-financial data basically includes the shareholders,the governance structure,the report of the board of supervisors,board reports and other important things,compared to the financial data,non-financial information often occupies most of the annual report,so it is very meaningful to accurately quantify non-financial information and analyze the rules behind it.Because the text information itself usually exists in semi-structured or unstructured form,it is not easy to quantify text information,but with the development of computer natural language processing technology,text mining technology provides a solution to these problemsThis article uses text mining technology to the annual report of China's listed banks for text mining,and uses the R language to quantify the text information in it,which makes the semi-structured or even disorganized text in order just like the digital information.According to the traditional taxonomy,this paper categorizes banks into four groups: the sample group of all commercial banks,the sample group of state-owned commercial banks,the sample group of joint-stock commercial banks and the sample group of urban commercial banks.After performing word frequency statistics of the above sample groups separately,the author classifies the high frequency words into the words with the positive color and the words with the negative color respectively according to the sentiment dictionary,then builds a regression model between words and listed bank performance to figure out the relationship between high-frequency words and the performance of listed banks.Through empirical analysis,this paper draws the following conclusions: Through the empirical analysis of the whole sample,this paper finds that negative words have an impact on the rate of return on equity of listed banks,and this kind of effect is negative,while the positive words in annual reports of the listed banks have no impact on it.2.Through the empirical analysis of the sample group of state-owned commercial banks,this paper founds that the negative words in annual reports of the listed banks have a negative impact on the rate of return on equity,and this effect is more significant than the full sample,while the positive words in annual reports of the listed banks have no impact on it.3.Through the empirical analysis of the sample group of joint-stock commercial banks,this paper finds that negative words have a negative impact on the rate of return on equity of joint-stock commercial banks,and positive terms have a positive impact on the rate of return on equity of joint-stock commercial banks.4.Through the empirical analysis of sample banks of city commercial banks,this paper finds that neither the positive words nor the negative words have obvious effects on the rate of return on equity.Here are also some suggestions at the end of the conclusions.