Research On Stock Average True Range Prediction With Web Media Analysis

With the rapid development of China’s Internet technology and the improving financial market mechanism,online public opinion gradually becomes an important factor affecting the stability of China’s stock market.Public opinions in the stock market include many aspects,such as fluctuations in the stock market,changes of investor sentiment and public discussions caused by fluctuations in the stock market.This paper proposed to use Average True Range index to measure the volatili ty of the stock market,and it mainly studies how to quantify public opinion by establishing the weighted emotion dictionary of the stock market with the help of two major platforms,CNKI and Baidu index,and how to construct an weighted emotion index to predict the Average True Range.How to accurately and emotionless reveal investor’s emotion is the key to establish the weighted emotion dictionary of the stock market.In addition,keywords in the dictionary should be highly relevant to the stock market.A t the same time,in order to achieve the dynamic monitoring of the stock market,keywords should be updated in time.Therefore,the specific steps are as follows:(1)Establishing initial keywords: grab more than 10,000 literature topics about "stock market" in the past three years by the “bazhuayu” software,conduct words segmentation with the help of the Python package “Jieba”,and establish the initial keywords data through Chinese words segmentation,filtering and word frequency analysis;(2)Keywords time difference method: setting y as the hs300 index of Average True Range data,x as a network search keyword TF-IDF values,use SPSS to calculate the time difference correlation coefficient of the daily data of the keyword in the initial keywords data and the hs300 Average True Range,then select leading keywords from the initial keywords data by the time difference analytic method;(3)Keyword importance ranking: use the random-forest algorithm in “sklearn” to rank keywords by their importance,and select the keywords with an explanatory ability of over 80% as the keywords for subsequent polar analysis.Then,the index of emotion weight was further constructed.The weight of each keyword was calculated by the logistic regression algorithm,and the polarit y analysis was carried out.Then,this paper will calculate the correlation between the constructed weighted emotion index and the Average True Range.The empirical study verifies the hypothesis that when investors pay more attention to keywords,the volatility of stock price will increase,which means,the Average True Range will increase.Finally,the last step is predicting the Average True Range of the next trading day of hs300.The Average True Range and the constructed weighted emotion index were inputted into the Recurrent Neural Network for prediction,and the parameters,lstm_size,max_epoch,num_layers,were adjusted for optimization.This paper quantifies public opinions by constructing a weighted emotion dictionary and weighted emotion index of public opinions.Then,it predicts the Average True Range,which provides certain theoretical and practical meaning for the construction of stock index prediction model and the application of deep neural network in the financial market...