AI in FIs: Learning Types and Functions of Machine Learning Algorithms

Through the lens of Financial Risk, this blog series will focus on Financial Institutions as a premier business use case for Artificial Intelligence and Machine Learning.

This blog series has covered how a financial institution (FI) can use machine learning (ML) and how these algorithms can augment existing methods for mitigating financial and non-financial risk. To tie it all together, the focus now will be on different learning types of ML algorithms:

  1. Supervised Learning
  2. Unsupervised Learning
  3. Semi-Supervised Learning

Deciding which learning type and ultimately which algorithm to use depends on two key factors: the data and the business use. In regards to data, there are two “formats” in which it exists. The first type is structured data. This type of data is organized and often takes the form of tables (columns and rows).  The second type is unstructured data. This type of data may have a structure of its own, but is not in a standardized format. Examples of these include PDFs, recorded voice, and video feed. This data can provide great value but will need to be reformatted so an algorithm can consume it.

 

Learning Types and Functions of MLA

 

Supervised learning algorithms draw inferences from input datasets that have a well-defined dependent, or target, variable. This is referred to as labeled data. Consider the scenario when an FI wants to predict loss due to fraud. For this they would need a labeled dataset containing historical transactions with a target variable that populates for a known fraudulent transaction. The FI might then use a decision tree to separate the data iteratively into branches to determine estimates for likelihood of fraud. Once the decision tree captures the relationships in the data, it can then be deployed to estimate the potential for future fraud cases.

Unsupervised learning algorithms draw inferences from input datasets with an undefined dependent variable. This is referred to as unlabeled data. These kinds of algorithms are typically used for pre-work to prepare data for another process. This work ranges from data preparation to data discovery and, at times, includes dimensionality reduction, categorization, and segmentation. Returning to our fraud example, consider the data set without the target variable (i.e., no fraud indicator). In this scenario, the FI could use an unsupervised learning algorithm to identify the most suspicious transactions through means of clustering.

Sometimes, a dataset will have both labeled and unlabeled observations, meaning a value for the target variable is known for a portion of the data. Data in this case can be used for semi-supervised learning, which is an iterative process that utilizes both supervised and unsupervised learning algorithms to complete a job. In our fraud example, a neural net may be used to predict likelihood of fraud based on the labeled data (supervised learning). The process can then use this model, along with a clustering algorithm (unsupervised learning), to assign a value to the fraud indicator for the most suspicious transactions in the unlabeled data.

 To learn more about ML algorithms and their applications for risk mitigation, please contact us or visit our Resources page for other ML and AI material, including the New Machinist Journal Vol. 1 – 5 .

 Hannah Wiser is an associate consultant with FRG. After graduating with her Master’s in Quantitative Economics and Econometrics from East Carolina University in 2019, she joined FRG and has worked on projects focusing on technical communication and data governance.

 

List of Terms to Know

AI in FIs: Introducing Machine Learning Algorithms for Risk

Through the lens of Financial Risk, this blog series will focus on Financial Institutions as a premier business use case for Artificial Intelligence and Machine Learning.

For any application of machine learning (ML) being considered for industry practice, the most important thing to remember is that business needs must drive the selection and design of the algorithm used for computation. A financial institution (FI) must be smart about which of these advanced tools are deployed to generate optimal value for the business. For many FIs, this “optimal value” can refer to one of two categories: increasing profitability or mitigating risk. In this post, we will focus on the uses cases for ML specifically related to risk.

Risk can be broken out between financial risk and nonfinancial risk. Financial risk involves uncertainty in investment or business that can result in monetary loss.  For example, when a homeowner defaults on a loan, this means the lender will lose some or all those funds.

Nonfinancial risk, on the other hand, is loss an FI experiences from consequences not rooted in financial initiatives. Certain events, such as negative news stories, may not be directly related to the financial side of the business but could deter potential customers and hirable talent. Some areas of risk may be considered either financial or nonfinancial risk, depending on the context.

When properly employed, ML enhances the capabilities of FIs to assess both their financial and nonfinancial risk in two ways. First, it enables skilled workers to do what they do best because they can off-load grunt work, such as cleaning data, to the machine. By deploying a tool to support existing (and cumbersome) business operations, the analyst has more time to focus on their specialty. Second, a machine has the technical capability to reveal nuance in the data that even a specialist would not be able to do alone. This supplements the analyst’s understanding of the data and enriches the data’s worth to the business.

The image below elaborates on the many kinds of risk managed by an FI, in addition to practical ways ML can supplement existing methods for risk mitigation.

More complex algorithms may do a better job of fitting the data, model at a higher capacity, or utilize non-traditional types of data (e.g., images, voice, and PDFs, etc.), but this all comes at a cost. The intricacies of implementing an ML algorithm, the commitment of time required to build a model (i.e., tuning hyperparameters can take days), and the management of unintended bias and overfitting render ML a considerable investment of resources. Not to mention, the robust requirements for computational power may require an FI to do some pre-work if a stable and capable infrastructure is not already in place.

As innovative as ML can be, any process will only be successful in industry if it produces value beyond its costs. Thanks to advances in computational power and available data, new approaches (e.g., neural nets) have broadened the universe of ML and its relevance, as well as better enabled traditional methods of ML (e.g., time series models). We will expand more on specific algorithms and risk mitigation use cases in later discussions.

Interested in reading more? Subscribe to the FRG blog to keep up with AI in FIs.

Hannah Wiser is an assistant consultant with FRG. After graduating with her Master’s in Quantitative Economics and Econometrics from East Carolina University in 2019, she joined FRG and has worked on projects focusing on technical communication and data governance.

 

 

AI in FIs: Foundations of Machine Learning in Financial Risk

Through the lens of Financial Risk, this blog series will focus on Financial Institutions as a premier business use case for Artificial Intelligence and Machine Learning.

Today, opportunities exist for professionals to delegate time-intensive, dense, and complex tasks to machines. Machine Learning (ML) has the ability to automate Artificial Intelligence (AI) and is becoming much more robust as technological advances ease and lessen resource constraints.

Financial Institutions (FI) are constantly under pressure to keep up with evolving technology and regulatory requirements. Compared to what has been used in the past, modern tools have become more user-friendly and flexible; they are also easily integrated with existing systems. This evolution is enabling advanced tools such as ML to regain relevance across industries, including finance.

So, how does ML work? Imagine someone is learning to throw a football. Over time, the to-be quarterback is trained to understand how to adjust the speed of the ball, the strength of the throw, and the path of trajectory to meet the expected routes of the receivers. In a similar way, machines are trained to perform a specific task, such as clustering, by means of an algorithm. Just as the quarterback is trained by a coach, a machine learns to perform a specific task from an ML algorithm. This expands the possibilities for ways technology can be used to add value to the business.

What does this mean for FIs? The benefit of ML is that value can be added in areas where efficiency, prediction, and accuracy are most critical.  To accomplish this, the company aligns these four components: data, applications, infrastructure, and business needs.

The level of data maturity of FIs determines their capacity for effectively utilizing both structured and unstructured data. A well-established data governance framework lays the foundation for proper use of data for a company. Once their structured data is effectively governed, sourced, analyzed, and managed, they can then employ more advanced tools such as ML to supplement their internal operations. Unstructured data can also be used, but the company must first harness the tools and computing power capable of handling it.

Many companies are turning to cloud computing for their business-as-usual processes and for deploying ML. There are options for hosting cloud computing either on-premises or with public cloud services, but these are a matter of preference. Either method provides scalable computing power, which is essential when using ML algorithms to unlock the potential value that massive amounts of data provides.

Interested in reading more? Subscribe to the FRG blog to keep up with AI in FIs.

Hannah Wiser is an assistant consultant with FRG. After graduating with her Master’s in Quantitative Economics and Econometrics from East Carolina University in 2019, she joined FRG and has worked on projects focusing on technical communication and data governance.

 

 

Avoiding Discrimination in Unstructured Data

An article published by the Wall Street Journal on Jan. 30, 2019  got me thinking about the challenges of using unstructured data in modeling. The article discusses how New York’s Department of Financial Services is allowing life insurers to use social media, as well as other nontraditional sources, to set premium rates. The crux: the data cannot unfairly discriminate.  

I finished the article with three questions on my mind. The first: How does a company convert unstructured data into something useful? The article mentions that insurers are leveraging public information – like motor vehicle records and bankruptcy documents – in addition to social media. Surely, though, this information is not in a structured format to facilitate querying and model builds.

Second: How does a company ensure the data is good quality? Quality here doesn’t only mean the data is clean and useful, it also means the data is complete and unbiased. A lot of effort will be required to take this information and make it model ready. Otherwise, the models will at best provide spurious output and at worst provide biased output.

The third: With all this data available what “new” modeling techniques can be leveraged? I suspect many people read that last sentence and thought AI. That is one option. However, the key is to make sure the model does not unfairly discriminate. Using a powerful machine learning algorithm right from the start might not be the best option. Just ask Amazon about its AI recruiting tool.[1]

The answers to these questions are not simple, and they do require a blend of technological aptitude and machine learning sophistication. Stay tuned for future blog posts as we provide answers to these questions.

 

[1] Amazon scraps secret AI recruiting tool that showed bias against women

 

Jonathan Leonardelli, FRM, Director of Business Analytics for the Financial Risk Group, leads the group responsible for model development, data science, documentation, testing, and training. He has over 15 years’ experience in the area of financial risk.

Real Time Learning: A Better Approach to Trader Surveillance

An often-heard question in any discussion of Machine Learning (ML) tools is maybe most obvious one: “So, how can we use them?”

The answer depends on the industry, but we think there are especially useful (and interesting) applications for the financial services sector. These consumers have historically been open to the ML concept but haven’t been quick to jump on some potential solutions to common problems.

Let’s look at risk management at the trading desk, for example. If you want to mitigate risk, you need to be able to identify it in advance—say, to insure your traders aren’t conducting out-of-market transactions or placing fictitious orders. The latest issue of the New Machinist Journal by Dr. Jimmie Lenz (available by clicking here) explains how. Trade Desk Surveillance is just one way that Machine Learning tools can help monitor a variety of activities that can cause grief for those tasked with risk management.

Would you like to read more about the possibilities ML can bring to financial services process settings? Download “Real Time Learning: A Better Approach to Trader Surveillance,” along with other issues of the New Machinist Journal, by visiting www.frgrisk.com/resources.

Introducing the New Machinist Journal

Who are the new machinists, and what are their tools?

The machinists of the 21st century are working with Artificial Intelligence (AI) and Machine Learning (ML), turning what has been science fiction into science fact. From learning algorithms that nudge us to buy more stuff to self-driving vehicles that “learn” the highways and byways to deliver us to our destinations safely, AI and ML are attracting considerable attention from a variety of industries.

FRG is currently researching and building machine learning proof-of-concepts to fully understand their practical applications. A new series, the New Machinist Journal, will explore in detail some of these applications in different environments and use cases. It will be published regularly on the FRG website. Volume 1, “What Artificial Intelligence and Machine Learning Solutions Offer,” is an overview of the subject, and is now available for download (click here to read it).

For more information, contact the FRG Research Institute, Research@frgrisk.com

Subscribe to our blog!