Technological applications are playing a more influential role in management in the contemporary business environment. Machine learning, artificial intelligence, and other algorithmic applications are some of the most common influencers in business applications. They present numerous solutions to business management problems, including banking risk management. In the last decade, risk management has gained greater prominence in financial services. In the past, banks focused on the detection, measuring, and reporting of risks. However, they are now leveraging on machine learning for greater accuracy and efficacy in risk management. As such, this paper explored different ways that machine learning applies in banking risk management. To achieve the objective of this study, the researcher conducted a comprehensive literature review on the topic of machine learning in banking risk management. The researcher found considerable industry and academic research focusing on developments in the financial services industry, especially in relation to risk management. It reviewed the literature, analysing and evaluating various risk management machine-learning techniques. It identified risk management problem areas and explored various ways of addressing them.
The review showed that machine learning learning in risk management in financial services sector was still under-researched. While there were many studies on credit risks, other risks such as liquidity risks, market risks, and operational risks saw minimal attention. Nevertheless, machine learning applications were found to have the potential to develop more effective risk management models. Machine learning is leveraged on different data types to predict potential events with greater accuracy and estimate losses associated with different risk types. In addition, the machine learning techniques in risk management were found to provide better and more accurate results than traditional statistical models. Though machine learning suggests improving banking risk management, there are some areas that need further study. For instance, the paper suggested in-depth studies on machine learning models for different types of banking risks.
Machine Learning in Banking Risk Management
Abstract
Technological applications are playing a more influential role in management in the contemporary business environment. Machine learning, artificial intelligence, and other algorithmic applications are some of the most common influencers in business applications. They present numerous solutions to business management problems, including banking risk management. In the last decade, risk management has gained greater prominence in financial services. In the past, banks focused on the detection, measuring, and reporting of risks. However, they are now leveraging on machine learning for greater accuracy and efficacy in risk management. As such, this paper explored different ways that machine learning applies in banking risk management. To achieve the objective of this study, the researcher conducted a comprehensive literature review on the topic of machine learning in banking risk management. The researcher found considerable industry and academic research focusing on developments in the financial services industry, especially in relation to risk management. It reviewed the literature, analysing and evaluating various risk management machine-learning techniques. It identified risk management problem areas and explored various ways of addressing them.
The review showed that machine learning learning in risk management in financial services sector was still under-researched. While there were many studies on credit risks, other risks such as liquidity risks, market risks, and operational risks saw minimal attention. Nevertheless, machine learning applications were found to have the potential to develop more effective risk management models. Machine learning is leveraged on different data types to predict potential events with greater accuracy and estimate losses associated with different risk types. In addition, the machine learning techniques in risk management were found to provide better and more accurate results than traditional statistical models. Though machine learning suggests improving banking risk management, there are some areas that need further study. For instance, the paper suggested in-depth studies on machine learning models for different types of banking risks.
Introduction
Machine learning technologies are inevitable in contemporary banking services. Specifically, they are integral parts of risk detection, measurement, reporting, and management (Financial Stability Board, 2017). As such, many studies now focus on the implications of risk management’s machine learning techniques in banking. For instance, Leo et al. (2019) indicated that banking risk management would be considerably different in the next decade. The expected changes in risk management approaches are attributed to new regulations, the evolution of customer expectations, and the emergence of new risks. In line with this argument, the application of new technologies and advanced analytics would impact product development, service provision, and risk management techniques. Leo et al. (2019) identified machine learning as one of the technological applications that would influence risk models. For example, machine learning can potentially improve the accuracy of risk models as they identify complex and non-linear patterns within the datasets. It also helps in the development of models with greater predictive power as they can integrate more information and be applied across many banking risk areas. Based on these arguments, this paper explored different ways that machine learning techniques are relevant to risk management in the banking sector.
Many studies suggest machine learning as a strategy assisting transformation of bank’s risk management functions. Usually, banks pursue higher returns, which can come with higher risks. These risks include interest rate risk, credit risk, operational risk, market risk, liquidity risk, and foreign exchange risk, among others. Managing these risks effectively will be vital to the overall bank’s performance. Financial Stability Board (2017) indicated that credit risk is the greatest risk that banks face. It was the greatest risk as it required the most capital to manage. Market risks largely arise from the bank’s trading operations, while operational risks are associated with losses due to external events and internal system failures. Besides, the calculation of regulatory capital to cover these risks will require banks to estimate their economic capital based on accurate models (Weirich, 2020). For instance, the model should integrate all the primary risks that banks face, such as liquidity, operational, credit, and market risks. Based on the suggested model, the bank’s risk management strategy will be successful in monitoring, managing, and measuring the risks. Consistent with this argument, Ala'raj and Abbod, 2016) indicated that machine learning is an important application in the design and implementation of the risk management model that incorporates all the banking risks. Therefore, this paper is important as it will provide in-depth insights into the extent to which machine learning supports risk management in financial services and identify areas for further improvements.
The paper assessed, analysed, and evaluated machine learning techniques and their applicability to management, assessment and monitoring bank’s risks. Moreover, it identified the challenges in management that remain underexplored and made suggestions for further studies. The paper leveraged on the existing literature to determine specific risks specific to banks and machine learning techniques. It evaluated applications of machine learning in risk management strategies of banks. The rest of the paper was divided into a discussion and a conclusion section. The discussion section presented an overview of risk management processes at the bank and provided a critical evaluation of machine learning applications in risk management techniques. The conclusion section summarises the findings in the paper and provides suggestions for further research and practice in the banking sector.
Discussion
Recent push to digitise financial services and new regulatory requirements has resulted in banks creating large amounts of unstructured data. These are high frequency data originating from consumer apps, metadata, client engagements, and other external data sources (Weirich, 2020). In addition, the desire to improve the analytical capabilities of banks and automate their business lines, including risk management, has resulted in powerful and analytical solutions. Due to the large amounts of data, machine learning has become inevitable in the development of financial service models. Hamori et al. (2018) showed that machine learning will improve insights into client preferences, monitoring operations, client support, client identification and risk management.
Machine learning encompasses computer science, engineering, and statistical tools that people and organisations use in diagnosing and addressing problems with specific systems and processes. Ala'raj and Abbod (2016) explained that machine learning delivers the tool or application’s capacity to identify concise patterns in data sets which can support decision-making processes. When banks are required to extract important information from data, they may instruct a programmer to provide an explicit and comprehensive execution process. While programmers may not always solve this problem, machine learning addresses such a challenge through endowment of programs with learning and adaptation ability. In line with this argument, Hamori et al. (2018) indicated that there are machine learning programs which can learn and improve, thus enhancing the capacity of an organization to diagnose a problem and ensure that it adapts appropriately. Yu et al. (2016) noted that machine learning tools are drivers of technological advancements such as search engines and other applications adopted within the financial services sector. Consistent with this argument, La Torre (2020) found that multiple technological developments were positively contributing towards the ability of the financial sector to explore and mine data, including financial data from customers and the market.
Many bank managers are adopting machine learning alongside algorithmic tools in the evaluation of complex relationships. Though machine learning is limited in its ability to determine causality, it creates cost minimisation opportunities, improved productivity and better management of risks (Yu et al., 2016). Also, banks are automating their operations to enhance the efficiency of their regulatory compliance. As such, banks are adopting machine algorithms as they do not rely on assumptions and focus on data and its distribution. They help in addressing complex non-linear relationships. For instance, credit scoring involves assigning a number to the client, suggesting the likelihood of them defaulting. Using machine learning tools, the bank can leverage on the many classification-related algorithms. These algorithms classify creditors and predict the probability of default (PD). They estimate loss given default (LGD) and exposure at default (EAD). The techniques help the bank’s risk management department develop models that can accurately predict and estimate PD, EAD, and LGD, thus supporting credit risk exposure estimation.
Machine learning techniques are preferable to traditional financial statistics in classifying and predicting accurate risk exposure. For example, support vector machine (SVM) is a reliable machine learning technique across several applications. La Torre (2020) described SVM as a reliable machine-learning algorithm utilised in credit scoring. Using this algorithm, a data item will refer to a specific point in a multi-dimensional space and every feature is given a value within a particular coordinate. The algorithm finds the hyper-plane. It can also detect the frontier separating different classes. Users can apply SVM with broader (<90 days past due) or narrower (>90 days past due) credit scoring scales (Ala'raj & Abbod, 2016). However, Bacham and Zhao (2017) found that credit scoring models using a broader scale had greater accuracy. As such, they allowed for improved prediction accuracy. However, the users of SVM must be careful as the methodology adopted may lead to non-random samples depending on the sample design and sample units’ behavior, which may result in sample selection biases. Since machine learning involves modeling based on learning from the existing data. This quality makes it vulnerable to challenges and biases affecting statistical approaches.
[...]
- Citar trabajo
- Mourine Atsien (Autor), 2022, Machine Learning in Banking Risk Management, Múnich, GRIN Verlag, https://www.grin.com/document/1326036
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X.