Predictors Of Realized Volatility By Investor Sentiment Based On Tribune Dictionary

At present,obtaining information from the Internet has become an indispensable channel for people’s life and work,and the advancement of software and hardware technology enhances development of using high-frequency data to study the realized volatility,of which the most extensive research is the volatility of the market index.At the same time,the number of Internet users continues to increase,and a large number of stock commentary texts have been posted on the Financial Forum website.People try to use these texts to predict the volatility of the broader market index and thus know about grasp the current market situation.Among the financial forum websites,the Oriental Wealth Internet Stock Bar Forum is the most well-known.It is difficult to improve the classification effect of the machine learning classification method used in the related research of the Oriental Fortune Internet Stock Bar Forum in the past,because there are many professional vocabulary and colloquial vocabulary in the text of the forum.Moreover,the processing level of noise processing and labeling tasks in the machine learning methods will also affect the final classification result.Therefore,from the perspective of lexicography,this paper aims to construct a new dictionary to capture the sentiment of financial forum texts.There are three kinds of emotion dictionaries used to analyze the financial field in the past.One is the basic emotion dictionary,the second is the dictionary constructed by the relative word frequency method,and the third is the translation of foreign financial dictionaries.This paper analyzes and finds that the above three dictionaries have their limitations,when applied to financial forums,they all cannot accurately identify text emotions.Therefore,this paper first uses crawler technology to crawl a large number of stock review texts from the “Shanghai Stock Exchange Index Bar” of the Oriental Fortune Network and then,combining the basic sentiment dictionary and the authoritative social media dictionary in the financial field,a financial forum sentiment dictionary is constructed by Sim BERT Base model,and the advantages of which are explored from the three dimensions of applicability,accuracy and comprehensiveness.Combined with the Financial Forum Sentiment Dictionary,a counting and indicator calculation program are developed to capture the forum posting sentiment,and then is constructed to a HAR-RV-EMO model to study the investor sentiment on the realized volatility forecast.The empirical results show that the fitting effect and prediction error of the HAR-RVEMO model with the investor sentiment of the forum are better than those of the original HAR-RV model.The daily mood fluctuation indicator of the forum has no significant impact on the volatility of the Shanghai Composite Index,which indicates that the daily indicator has a delayed effect,people will not directly affect the day’s operation because of a certain day’s abnormal mood.However,the weekly and monthly indicators have a significant impact,and the monthly sentiment indicator has a greater impact than the weekly indicators,which indicate that investors would carefully consider this month’s sentiment changes when making decisions.When monthly sentiment is more consistent but weekly sentiment is more inconsistent,the stock market is prone to greater volatility the next day.In addition,the prediction effect of the HAR-RV-EMO model is tested from both intuitive and objective aspects.First of all,through the images of the predicted value and the actual value in the sample,it can be found that the model can respond quickly to changes in investor sentiment and can better predict the volatility of the Shanghai Composite Index.Secondly,the extra-sample daily fluctuation prediction method of rolling time window is used to calculate the predicted values of HAR-RV and HAR-RV-EMO wave prediction models,and 2000 Bootstrap simulations are used to perform SPA test by using five loss functions.The SPA test results show that compared with the HAR-RV model,the HAR-RVEMO model has better ability in volatility prediction,that is,the inclusion of forum investor sentiment consistency index can improve the prediction ability of the volatility model.