A Research On Relationship Between Different Social Media Platforms And Stock Price

Recently,the social media is increasingly getting involved into every aspects of the public's daily life.Meanwhile,as the continuous development of big data and artificial intelligence,researchers are now having the ability to realize their ideas to track and predict the users' behavior on the social media platforms.many researches have started predicting the behavior of financial stock market so called as a very complicated system using the public's sentiment data.Then according to their research they have established the models of predicting the stock price or some other market indices.All these related results of researches have shown that the combination of public emotions and market data could possibly increase the accuracy of predicting this complecated market.However,among all these updated researches,there are not many research reports are able to give a good comparison of the difference of predicting stock prices according to different group mood data.Also,they did not draw a well conclusion to show how this difference can affect the accuracy of predicting.Because of the totally different distribution of the major user groups,obviously,different styled user textual data are being made by those different type of users.For example,the sina weibo platform is famous for its broad public topics and the xueqiu is focused on investment and stock analysis.Users from these two different platforms would take totally different style of discussion of certain stocks.This thesis pays attention to difference between social media platforms,especially targeting two big public social media platforms of xueqiu and sina weibo.It mainly performs a comparison of the relationship of the public's sentiment change from different platforms and the stock price movement.In this research,(1)users' comment data were scraped and collected from weibo and xueqiu representing the sentiment status of the users from these two platform.(2)After the data preparation phase,including data processing,data cleaning and data classification,we deployed bag-of-words technique in sentiment analysis to the different data set using a financial field related dictionary so that the emotional scores by time of user groups from different platforms are generated.This sentiment time series are also applied with a POMS processing to make it smooth.(3)Then,analysis has been taken to measure the difference between the xueqiu users' sentiment and weibo users' sentiment with certain stocks movement.Granger causality analysis is used to validate the causality of sentiment time series generated by xueqiu and weibo with the financial time serieses.It is found that,under certain lag time conditions,the mood series of xueqiu users has stronger causality with stock price time series than those of sina weibo users.It has more significant statistical features.(4)Finally,based on the sentiment time series both of xueqiu and weibo,the SOFNN model is utilized to build the stock price movement prediction model.This model takes the mood time series and stock price change as the inputs,then outputs the price change prediction value.It has a relatively good performance on the stock price change prediction.In the criterion of MAPE and direction accuracy,the prediction model which takes the xueqiu users' sentiment time series as the input gives better performance than that whose inputs of the weibo users' moods.The model has a mean average percentage error of 1.83% and a direction accuracy of 72.3%.