Can Web News Media Sentiments Improve Stock Trading signal Prediction?

The technical analysis and machine learning have been integrated in stock trading signal forecasting. And it has been proved that there are some weaknesses in technical analysis because of the complex environment in the stock market. In our prediction system, web news media sentiment analysis is regarded as a supplementary way to cover the shortage of technical analysis. It is considered to bring the stock market sentiment which reflects the subjective information of investors into the prediction system. Web news media sentiment indicators (WNMS) are designed to bring the information about stock market sentiment in our system. The WNMS is generated by analyzing the variance of sentiment elements from the news in the Stock Timely Rain Sector of Sina Finance and Economics Website and it is imported into the prediction system as features combined with common feature indicators (CFI). GMKL is applied to establish the relationship between the trading signals generated by piecewise linear representation (PLR) and the features of the trading signals (SCFI). Comparative experiments are adopted in nine stocks from Shanghai and Shenzhen Stock Exchange to determine the effect of PLR-GMKL and WNMS in prediction. From the aspects of the prediction accuracy and the profit, the final comparative results show that the PLR-GMKL model performs better than the PLR-WSVM model. And the prediction system performs best when adding WNMS into features and using PLRGMKL model.