An Empirical Analysis Of The Impact Of Internet Financial News On China's Stock Market

Currently,the global economy is in an important period of transformation.People's demand for how to obtain information in a timely and accurate way has become stronger.At present,Internet news media have become the main source of information for the general public because of its convenience and timeliness.The stock market is endlessly changing.No matter which reports may cause changes in the stock market.How to quickly and accurately analyze the stock market according to news reports on the Internet has become a problem that financial institutions and investors have to consider.In recent years,due to the rise of technologies such as data mining and artificial intelligence,research ideas for studying the laws of the stock market through Internet media information have become feasible.But because this kind of research involves many disciplines such as psychology,economics,computer science,statistics,and linguistics,this kind of research is still in its infancy.The main research objects of foreign scholars in this field are English texts,and some related research tools have been developed.However,due to the different conditions of language,media environment,and technical means,Chinese word segmentation technology and sentiment analysis methods are not mature enough.Therefore,there are still many technical problems to be solved in this field.This shows that exploring the impact of Internet financial news on China's stock market is a research topic that has both theoretical significance and practical value.This article explores the impact of financial news on China's stock market from the perspective of sentiment and topic of Internet financial news,and hopes to provide some new ideas for future research in this field.In the theoretical and methodological research section,this article first introduces the relevant theories of text mining technology,including Chinese word segmentation,feature extraction,text classification,and so on.Then it expounds the sentiment analysis of text and introduces how to carry out sentiment analysis and the construction method of sentiment dictionary.Next,it explains the extraction method of financial news topic information,that is,the topic model,and introduces the LDA topic model in detail.Finally,the models used in this paper to forecast the stock market,namely Logistic Regression and Support Vector Machine,are introduced,and the model evaluation indicators used in this paper are described.In the empirical analysis section,this article studies the impact of Internet financial news on the stock market from the following two aspects.Firstly,this article studies the relationship between financial news and stock prices from the perspective of sentiment analysis.The stocks selected here are 4 stocks: Ping An,Guizhou Maotai,China Merchants Bank and China Unicom.Before analyzing,it is necessary to quantify the sentiment tendency of news.This article uses the method based on semantic rules to analyze the sentiment of news.First,build sentiment dictionaries required for this research.With the help of the existing basic sentiment dictionaries and the SO-PMI algorithm,the basic dictionaries are extended to obtain the sentiment dictionary for the financial news field,as well as the degree word dictionary,the negative dictionary,and the transition summary dictionary.Then we set corresponding weights for the words in each dictionary.Then,based on the sentiment dictionary,a series of semantic rules are defined to calculate the sentiment value of the sentiment unit,and then the sentiment value of the sentence and the text are calculated.Next,we perform a preliminary analysis on the correlation between the sentiment value series and the individual stock return series,and it was found that the correlation between the two for the four stocks will change over time.And a few days after the news release,the sentiment tendency of the news still has a certain degree of influence on the return of individual stocks.Finally,this paper uses Logistic Regression and Support Vector Machine to build models to predict the stock price trend,and compares the prediction performance of the experimental model and the three benchmark models.The empirical results show that no matter whether Logistic Regression or Support Vector Machine is used,the prediction accuracy of the experimental model is higher than the three benchmark models for the four stocks,which validates the method proposed in this paper to quantify news sentiment based on the sentiment dictionary and semantic rules.In addition,on this issue,the prediction performance using Support Vector Machine is better than Logistic Regression.Finally,this article compares the models based only on historical closing prices and based on historical closing prices and news sentiment values.The result shows that the introduction of news sentiment values based on historical stock price data can indeed greatly improve the prediction performance of stock price trends.Secondly,this article studies the relationship between financial news and industry sector indexes from the perspective of the topic.Here we take nine industry sector indexes in Shanghai Stock Exchange as the research object.First,the news text is preprocessed and a news corpus is formed.Through the LDA topic model,the news corpus was clustered into 80 topics,and the topic probability distribution of each news was obtained.The characteristics of the news topics and the topic distribution were briefly analyzed.Then we use Kendall rank correlation test to test the correlation between rankings of daily financial news topic distribution and rankings of industry sector return.The result shows that there is a significant correlation between the two,indicating that the topic distribution of news has an impact on the industry sector return.Finally,this article takes three representative industry sectors as examples,and uses Granger causality test to analyze the time series of sector index return and the corresponding topic distribution.Besides,the trend of the three representative industry sector indexes was predicted using the topic distribution of financial news.The predicted F1 values are 69.7%,69.69%,and 62.3%,indicating that news topics help predict the trend of the industry sector index.Because the distribution of news topics represents the media's attention to different topics,this reflects the current events and policy changes during the period,which in turn will affect the decision-making behavior of investors and have an impact on the stock market.This empirical result validates the claim that news topics have an indirect impact on the stock market.