The closure of an incompetent bank by a state or federal authority is a bank failure. Closing national banks is within the purview of the comptroller of the currency; closing state-chartered banks falls under the ambit of the relevant state’s banking commissioner. When a bank cannot fulfill its responsibilities to its customers and other parties, it must close Leisure, M (2022). Bank failures can occur for a variety of causes. These are a few of the hazards, fluctuating interest rates, improper management techniques, lax accounting practices, and growing competition from non-depository organizations. Since the crisis, reducing the magnitude of deposit insurance liabilities has become a top priority for bank regulators. Even the idea that closing banks before becoming financially distressed are the optimum regulatory strategy has been floated. To prevent another severe financial crisis, it is crucial to identify potentially failing banks as soon as feasible.
Numerous technologies have been created during the past thirty years to identify issue banks. Recent improvements are based on neural and fuzzy logic networks, as opposed to earlier models that mainly used statistical techniques [1], [4], [6]. The majority of these algorithms categorize data as either bankruptcy-related or not. However, in the actual world, banks are classified according to how likely they are to file for bankruptcy Leisure, M (2022). A system of early warning that can “flag” possibly failing banks is what regulators need. Once such banks have been located, various preventive programs tailored to each bank’s requirements can be implemented, helping to prevent a significant banking failure.
In addition, over the previous 30 years, cluster analysis has split various objects into groups or clusters to maximize the degree of association between two things when they are members of the same set and minimize it when they are not. Robert Tryon coined “cluster analysis” more than 80 years ago. Since then, cluster analysis has been extensively used in various industries, including astronomy, medicine, and archeology. No predetermined classes are used in clustering; instead, objects are solely clustered together based on their similarity Andersen and Jensen, (2021). Both dependent and independent variables are treated equally in Copyright The Second World Congress on Autonomous Applications and Systems. Then when clusters are discovered, the user must assess their significance. There are three main approaches that cluster analysis uses. These are supported by neural networks, fuzzy logic, and statistics. We shall use a self-organizing neural network in this case study. In this study, we use cluster analysis to “flag” banks that may be in trouble.
However, our objective is to group banks based on their financial statistics. The Federal Deposit Insurance Corporation’s annual reports can be used to access the information (FDIC). The United States Congress established the FDIC as a separate entity. It administers receiverships, inspects and regulates financial institutions, and guarantees savings. Andersen and Jensen (2021). Administrators employ the CAMELS (Capital adequacy, Asset, Management, Earnings, Liquidity, and Sensitivity to Market Risk) grading system to evaluate a bank’s overall financial health. 90 00 banks in the US have received CAMELS ratings. The United States government utilized it to choose banks for the 2008-2009 capitalization schemes.
Consequently, during the study case, we chose a hundred organizations and got financial information about them from the FDIC yearly survey for the previous year. Based on the CAMELS system, we modify the following ratings:
- NITA is the product of net income and total assets. NITA stands for net income after tax. The NITA values of failing banks are extremely low or even negative
- NLLAA is the ratio of adjusted assets to net loan losses. Total debts are subtracted from all assets to calculate adjusted assets. Typically, NLLAA levels for failing banks are higher than those for healthy banks.
- NPLTA is the non-performing loans as a percentage of total assets. Loans that are 90 days or more past due, as well as non-accrual loans, fall under non-performing loans Haralayya, D (2021). The NPLTA values of failing banks are typically more significant than those of healthy institutions.
A handful of banks may be having some financial issues, preliminary analyses of the statistical data suggest. We should be able to discover groups of banks with similar problems using clustering Haralayya, D (2021). Since there are typically fewer clusters identified by a SOM than neurons in the Kohonen layer, several input vectors drawn to nearby neurons may represent the same cluster.
Far from that, we require a test set to test a neural network, including a SOM. We may get a list of failed banks and pertinent financial statement information from the FDIC Annual Report. According to several pieces of research on bank failures, failing banks can be identified between six and twelve months before the call date and, in some cases, even four years earlier. Although solvency and liquidity are the most crucial indicators of failure in the immediate aftermath of the call date, asset quality, profitability, and management strategies become more vital as time passes Bellia et al. (2021). We gathered the financial statement data from 10 banks that failed the previous year to test the SOM performance. The mean, median, and standard deviation (STD) values of their CAMELS ratings are shown in Table II. Ten input vectors can now be used to observe the SOM reaction.
Moreover, a word of warning. Although a SOM is an effective aggregating tool, the precise significance of each cluster is not always obvious, necessitating the assistance of a domain expert to interpret the findings. A SOM is a neural network as well, and the quality of any neural network depends on the data it is fed. We have only used five financial factors in this case study Bellia et al. (2021). Although academics in the field employ up to 29 economic variables based on the CAMELS rating system, we may need many more variables that store additional information about bank performance to identify problem banks before they fail. The SOM was then trained using the data. There were five clusters found. One of the clusters caught my attention in particular. It included clients who had taken out home equity loans. These clients were in their forties, married, and had young children.
In conclusion, one of the clusters caught my attention in particular. It included clients who had taken out home equity loans. These clients were in their forties, married, and had young children. The bank assumed they were taking out loans to cover their children’s college expenses. As a result, the bank set up a marketing campaign to promote home equity loans as a way for people to pay for college Abbas et al. (2021). However, the campaign’s outcomes were dismal. Further examination revealed that the SOM picked out the SOM’s interpretation as the issue.
As a result, the bank provided additional details about its clients, including the types of accounts, deposit methods, and credit card systems. After the SOM was retrained, it was revealed that borrowers of home equity loans were also in their forties. A financial crisis was sparked in 2008 by several bank collapses. However, although massive central banks and government intervention aided economic recovery, the world’s financial institutions are still in danger. Therefore, it’s crucial to spot failing banks as soon as feasible Abbas et al. (2021). The immense potential of a self-organizing neural network as a tool for completing this task has been amply illustrated in this research. The findings demonstrate that self-organizing maps can carry out job clustering and identify banks needing prompt regulatory intervention.
References
LESCURE, M. (2022). GENERAL INTRODUCTION Why do some banks fail to fail?
Andersen, T. B., & Jensen, P. S. (2022). Too Big to Fail and Moral Hazard: Evidence from an Epoch of Unregulated Commercial Banking. IMF Economic Review, 1-23.
Haralayya, D. (2021). Study on Non-Performing Assets of Public Sector Banks. Iconic Research And Engineering Journals (IRE), 4(12), 52-61.
Bellia, M., Maccaferri, S., & Schich, S. (2021). Limiting too-big-to-fail: market reactions to policy announcements and actions. Journal of Banking Regulation, 1-22.
Abbas, F., Ali, S., Yousaf, I., & Wong, W. K. (2021). Dynamics of funding liquidity and risk-taking: evidence from commercial banks. Journal of Risk and Financial Management, 14(6), 281.
write