Stock Price Prediction Base On Network Public Opinion And Support Vector Machine

Since the establishment of the stock market, the stock as the coexistence of high-yield and high-risk, has attracted a lot of investors. With the rapid development of the internet platform and the arrival of big data era, stock traditional technical indicators can’t be met people’s demand analysis and forecasting stock prices.In order to improve the prediction accuracy of the stock price, based on network public opinion and technical indicators, a stock prediction model (NPO-SVM) is put forward by support vector machine regression. Firstly, stock comments such as micro-blog and stock forum are divided into positive, negative and flat categories, and stock network public opinion is defined by the ratio of positive comments. Then, the input data of stock network public opinion and technical indicators with correlation coefficient above0.6are reduced to several factors by the principal component analysis. Finally, the model is built with the reserved factors by support vector machine regression. Compare with the support vector machine model based on technical indicators (TI-SVM) and the support vector machine model based on empirical mode decomposition, the NPO-SVM has higher prediction accuracy which can provide a reliable method of predicting stock prices for investors. The main research work of this thesis can be summarized as follows:(1) A new method to achieve a text message sentiment classification is put forward by machine learning which accuracy is85.4percent. The correlation coefficient between stock network public opinion and stock price is0.7. It has a strong correlation between the stock price and the stock network public opinion.(2) By using support vector machine regression model based on network public opinion and stock technical indicators, the stock price can be predicted. The simulation results show that the maximum relative error of NPO-SVM was2.7percent, the average absolute error was0.092, the average relative error was0.7percent and the trend accuracy rate was76.37percent. Compared with the model TI-SVM and EMD-SVM, the NPO-SVM is a higher efficient stock prediction model.