Research On China Stock Market Short-term Price Forecasting Based On Sentiment Analysis

With the economic development in China over the past few years,people have accumulated more and more wealth.Traditional saving can no longer satisfy people’s increasing demand for investment,while the stock market,as a kind of important investment channel and thanks to its attracting rate of return,has been more and more popular with investors.The prediction of share price has thus become a hot issue paid attention to by investors.For the traditional way of thinking for prediction of share price,the change of share price is predicted by extracting patterns from data of historical price of shares.However,change of share prices is caused by various factors.In particular,in domestic market,given that investors behave irrationally,they just follow what other people are doing.Under this context,it is of great value to study the response of investors to news events and their inclination in order to enhance the performance of share prediction model.In this thesis,a short-term share price prediction model is establishing,taking in to account the analysis of investors’ inclination.Based on this,a share transaction strategy suitable for short-term investment is put forward.First,regarding a complicated and large amount of shares in the share market,the historical price data set and corresponding stock comment data set are established by screening indicators and individual stocks that can represent the market situation.LSTM prediction model based on the historical price data is established and compared with ARIMA,classical time sequence-based prediction model.Second,sentiment analysis of stock comments is studied by using SVM model to categorize the comments based on sentiment,and investor emotion indicators are established.Then both the investor emotion indicators and historical share price data are used as the input of LSTM to build a share price prediction model integrating both the investor emotion indicators and historical transaction information.According to the experiment results,the model is excellent in terms of data set,outperforming LSTM model and ARIMA model which only focus on the historical share price data.Last,with the above prediction model based on sentiment analysis,taking into traditional investment analysis technologies,the short-term share transaction technologies are studied,and a set of short-term share transaction strategy is developed,integrating initiative selection of shares,timing and position management.Such strategy is examined through firm offer simulation in Shanghai and Shenzhen stock markets.The experiment shows that the method of short-term prediction of share price based on sentiment analysis is feasible and effective for given samples,and the transaction strategy developed in the thesis is effective too,thus providing a feasible new way of thinking for a wide range of investors to invest in the share markets in a more rational manner.