Predicting the Effects of News on the Financial Market Based on Machine Learning Technology

Predicting the impact of news on the financial market is a complex task, and even the most competent companies providing analytical services in the stock market field can only partially forecast changes there. Recent studies in machine learning have allowed many scientists to try new methods for setting new objectives. The question of testing machine learning methods for analysing the impact of news on the financial market based on FinancialPhraseBank data is further investigated in our work. The procedures and means of function analysis of the financial market and the impact of news on it is the object of the research. Their analysis enables them to predict how a particular information (positive/negative/neutral) affects the financial market. The most challenging issue is finding the news's semantic connection with the corresponding financial market changes to predict their value. Considering the methods of implementing the main functions, we have applied such machine learning methods as naive Bayesian classifier, logistic regression, and the support vector method. The corresponding software is developed based on the outcomes, and its structure is described. The logistic regression method copes best with analysing the news's impact on the financial market and shows an accuracy of 75.67%. This is not a desired result, though it is the most significant indicator. The support vector method performs worse in accomplishing the task and shows an accuracy of 72.78
^%
, slightly lower than the one obtained by the logistic regression method. And the naive Bayesian classifier method shows the lowest score, and its accuracy is only 71.13% which is less than the two previous methods.