A Novel Machine Learning Based Method for Stock Price Crash Prediction

Abstract

Corporate Governance is a very well-known and accepted technique for the assessment of company performance in the stock market and to predict and ensure that the company’s stock value will not fall. This practice throughout the globe helps ensure the safety of investor’s money and also keeps company’s stakeholders and shareholders on board with the truth about the real strength of the company and its worth. The board of directors in the corporate governance keeps transparency between the managers and the owners of the company. However, it has been seen that the composition and characteristics of this board affect the overall performance of the company. Therefore, in order to assess the performance of this board of governors/directors in light of the characteristics and composition of board, I have implemented a new technique of machine learning that can assess if the company’s stock value will crash in the stock market or not, depending upon the characteristics and composition of the board. This thesis uses data from Bloomberg Platform, Osiris and Corporate Library covering 500 banks and financial institutions to validate our algorithm. Moreover, in this thesis, I have compared the empirical results of this algorithm with the baseline known algorithms of SVM and logistic regression. Results show that the proposed algorithm is more accurate than the baseline methods. The thesis concludes with the effects and role of corporate governance features in stock price crash prediction.