Research On Investor Sentiment Analysis Based On Text Mining And Its Multi-fractal Cross-correlation With Media Attention

In recent years,with the rapid development of the Internet,the Internet has become a platform for information,communication and comment.It is precisely because of the openness of network information and resource sharing that the cost of investors' access to financial market information is gradually reduced,and the impact of information asymmetry is gradually reduced,but at the same time,because Chinese stock market is not fully mature,Chinese investors are irrational.The degree is high,and its emotions are easily affected by the attention and subjective views transmitted by the Internet media,which in turn leads to changes in investor investment decisions and investment behaviors.At present,scholars at home and abroad mainly study the influence of media or investors on the stock market from a single perspective,but lack of discussion on the relationship between the two.Studying the relationship between media attention and investors has important implications for studying the interactions between the two and understanding their relationship to financial markets.Because the financial market has a complex fractal structure,the Multifractal Cross Correlation Analysis Method(MF-DCCA)can describe complex systems in detail,which can explain the multifractal features between two time series and the local features of complex systems.Accurate portrayal.Therefore,the paper uses MF-DCCA method to analyze the correlation between media attention and investor sentiment intensity,and describes and analyzes their fractal features.This is a new idea for the application of the MF-DCCA method in the financial field.In order to more accurately analyze the relationship between media attention and investor sentiment intensity,the thesis firstly fails to accurately calculate the investor's sentiment intensity based on the sentiment analysis method based on traditional sentiment dictionary,and proposes an improved sentiment analysis method.This method uses the Word2 vec model to extend the Hownet sentiment dictionary.At the same time,in order to make the calculation result more objective,the Word2 vec model and the vector normalization algorithm are used to redefine and calculate the degree adverb weight.Then the paper takes China's A-share market as an example,selects 10 stocks in five industry sectors as research samples,and uses MF-DCCA method to empirically study the correlation between media attention and investor sentiment and attitude of these 10 stocks.The Baidu Media Index is used as a proxy variable for media attention.The Sina Finance website is used as a data source for calculating investor sentiment intensity,and emotional calculation is performed through improved emotional intensity calculation method.The experimental results show that the construction of the emotional dictionary and the redefinition of the degree adverb weights effectively improve the accuracy of emotional intensity calculation.Through empirical research,it is found that there is a correlation between the media attention of 10 stocks and the investor's emotional intensity,and it has obvious multi-fractal characteristics.In addition,there are multiple fractal features between media attention and investor sentiment intensity.difference.This study is of great significance for a better understanding of the nature and laws of investor sentiment changes.