Discussion on the feasibility of news sentiment analysis to predict stock prices

Abstract

Since its birth, China's A-share market has been regarded as a “policy city”, which means that the introduction of policies will cause earth-shaking changes in the overall market, and it will be less affected by the overall public opinion. In order to verify the influence of public opinion on the overall market index and the overall trend of individual stocks, it is proposed to use a linear regression (Linear Regression) and polynomial regression model (Polynomial Regression) based on the widely agreed academic community to fit the overall market news sentiment tendency and the magnitude of the corresponding individual stock and market index changes. This model first collects the corresponding news data of Resset database, and uses the natural language processing pre-training model to quantify the corresponding news headline data into sentiment analysis values, and then enters the sentiment analysis values together with the stock price, stock price change range and stock index and stock index change range of the corresponding date into Excel to fit the corresponding data. Comparing the R2 of the corresponding regression models of different stocks and stock indices, the regression results are verified and compared to verify the feasibility of relevant news sentiment analysis for stock price prediction. The experimental results show that the correlation coefficient R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> between the sentiment analysis value and the corresponding stock price or index fluctuation value of the polynomial regression model is less than 0.2, and the correlation is not enough to reach a significant level, and other indicators need to be introduced to achieve feasibility.