| With the rapid development of modern market economy,there are also higher risks.As one of the typical high-risk and high-income representatives,the stock market has become the focus of investors’ research.However,how to use various factors of market reaction in the prediction and analysis of stock price has become a key problem.Therefore,this paper focuses on several stock price prediction models after integrating micro blog sentiment analysis The following studies were conducted:First of all,the index input to the model is screened,referring to the selection of relevant stock forecasting research on stock index,this paper selects 40 indexes as the initial index,then uses R cluster analysis and the correlation between the index and stock price to screen again,and uses the method of principal component analysis to reduce the dimension to get five principal component factors.Secondly,the micro blog text data obtained by crawler technology will be cleaned,and the API interface in Baidu NPL with better emotional tendency analysis will be selected to build the emotional analysis model,and the micro blog text will be analyzed and calculated to get the emotional value SA_Value,at the same time,the paper analyzes the causality between the emotional analysis of micro-blog and the stock price.Finally,this paper selects the logistic model,SVM model and LSTM model of statistics,machine learning and deep learning respectively to establish the fusion SA_SA logistic model,SA-SVM model and sa-lstm model are used to predict and analyze the stock price after value.The empirical object of this paper is two stocks with high proportion weight in Shanghai Stock Exchange 50 and different industries: Ping An(601318)and Moutai(600519).The experimental results show that the error rates of logistic,SVM and LSTM models are 33.9%,8.4746% and 0.6874%.The analysis shows that the relative error rate of LSTM model is 1 lower than that of logistic and SVM 5-50 times,the results show that the LSTM model in deep learning has better nonlinear fitting ability in stock price prediction,and can make better use of trading technical indicators to make a certain trend judgment on the stock market;moreover,the prediction results of SA logistic,SA-SVM and sa-lstm models are 30.5%,6.7797% and 0.6874%,respectively.Compared with the data of Maotai in Guizhou,it can also be found that the introduction of Micro blog emotional value can reduce the prediction error rate to a certain extent.Finally,this paper applies the forecast results to the test samples for simulation trading.The trading results show that under the six models,the profit rates of Ping An are:-3.59%,-1.70%,3.05%,-3.67%,-0.23% and 8.56%,respectively,and the profit rates of Maotai in Guizhou are:-3.27%,9.08%,4.21%,-5.81%,12.07% and 6.43%.It is found that the SA-SVM model and sa-lstm not only have smaller relative error in prediction,but also have relatively high profit in transaction.It shows that the prediction method has good robustness and good adaptability to data jitter.At the same time,the integration of micro blog sentiment analysis and prediction model can monitor the stock market more effectively. |
