Essays on deep learning and complex systems theory for financial industry

Abstract

With the advent of the big data era, artificial intelligence technology is rapidly developing, and efforts to apply it to the financial industry are increasing. However, existing studies to introduce artificial intelligence into the financial industry have mainly focused on the development of artificial intelligence algorithms, and research on the nature of the financial system, complex systems, is insufficient. Therefore, this dissertation proposes a methodology that utilizes the complex system of the financial system and grafts it with the AI methodology. Specifically, in the first essay, the causal relationship of stock prices is identified using the complex system theory, a network is created based on this, and stock prices are predicted using media information (online news). This essay proves that the method using the network has higher performance than simply predicting stock prices using the media information of each company. In the second essay, the causal relationship of stock prices is identified using the complex system theory, a network is created by removing the noise of stock prices, and global financial indices are predicted by deep learning with this technical analysis. This essay proves that predicting financial indices based on the generated network shows higher predictive power than existing studies. In the third essay, the default risk is predicted through the production network (SCM) by changing the existing complex system theory, and the default risk of each company is predicted through financial statements using artificial intelligence methodology. In this essay, a method to integrate the two methods was proposed, which showed high accuracy, precision, and recall compared to the existing methods. This thesis not only contributes academically in that it grafts artificial intelligence and complex system theory to the financial industry but is also practical in terms of developing an algorithm that can be applied in the financial industry by checking the results based on real data.