Analyzing effect of news polarity on stock market prediction: a machine learning approach

In finance, the stock market and its trends are volatile in nature. In the stock market, which is dynamic, complex, nonlinear and non-parametric, accurate forecasting is crucial for trading strategy. This need attracted researchers to detect fluctuations and to predict the next move. It is assumed that news articles affect the stock market. In this work, non-measurable data like financial news headlines has been transferred into the measurable data. We investigated the relationship between news and their impact on stock prices. To show this relationship, we applied the sentiment analysis data and the price difference between the day before the news was published and the day of the news to the classic machine learning models such as SVR, BayesianRidge, LASSO, Decision tree and Random forest. The observations showed that SVM performs well in all tests. The prediction error in this model is 0.28, which is much less than that of the random news tagging. Also based on our tests, using a computer for tagging is as good as manual tagging.