Abstract:
Banking sector is principally responsible for holding financial assets/resources in any country’s economy. Therefore, Bank Failure has far greater impact on the overall economy of a country compared to any other business. It can rapidly pour out to other banks and financial institutions and therefore has an avalanche effect. In order to evade fateful financial scenarios, rigorous regulations have been put in place along with technology to monitor, track and forecast critical financial parameters. Numerous statistical techniques and machine learning approaches have been widely employed for pre-emptive decision making to preclude the potential financial crisis. Banks employ domain experts, who exploit their expertise along with these tools for financial performance assessment of the financial institutions. These experts, based on performance assessment, recommend actions to prevent bank failure. Hiring domain experts exhaust substantial financial resources. Moreover, in spite of financial burden, the recommended actions do not suffice to cease bank failure most of the time because; it is very hard to generalize all the knowledge due to complex correlations of the financial parameters. The success of Artificial Intelligence (AI) across different domains attracted financial institutions to adopt much powerful AI methods to replace inefficient old methods. In an effort to employ AI for assistance in financial decision making, this research work proposes a novel deep recurrent neural network for bank failure prediction. In this work, we propose a two-layer recurrent network with Long Short-Term Memory (LSTM) cells. To validate the proposed algorithm, we collected data of 5946 banks from United States in the time span from 2004 to 2018. In total we have collected 43 financial ratios/variables over fifteen years for each of the bank. The performance of the proposed algorithm is compared against that of widely adapted SVM and Logistic Regression methods. The results vindicate the superiority of our proposed approach. The thesis work concludes with a comprehensive study of effect and role of different parameters towards bank failure.
