Strengthen AML Compliance With Data Mining
To build an effective anti-money laundering (AML) compliance program, you must first start with data you can trust. But in the age of big data we are overloaded with streams of data from various sources—and we often lack the ability to derive insights from it.
Today, data mining is a critical component of any compliance program. By helping identify hidden patterns, discover unknown relationships in your data, and predict behaviors and trends, data mining can be a strategic part of your business.
In this webinar, we examine data mining techniques that can be used to identify risk factors in your compliance program, monitor customer activity and provide insights into your overall business.
Here is a summary of what is covered during the presentation:
What is data mining?
Data mining is a fairly new concept as far as how to use information to help us make decisions. In this case the information we’re relying on is data that we already have within our own organizations.
Usually when trying to use data to solve a problem, the data that we want tends to be disjointed; it can be in various places—for example, in customer relationship management (CRM), human resource management (HRM) and enterprise resource planning (ERP) systems, as well as in application software or Excel—and that can be challenging.
Getting all of the information needed can be tedious and time consuming, and other cases of fraud could be happening while you’re trying to gather all of the information you need to conduct a proper investigation.
With data mining, you can take all of the information from these various sources and combine them together in such a way that you not only have easy access to the information, but you’re able to quickly learn from it and gain new insights through the use of data models.
Today we’re experiencing a phenomenon that we’re going to call the “data explosion” problem.
Over the last two decades, technology has advanced rapidly, allowing us to generate so much data that we don’t know what to do with it all.
We’ve become experts in collecting data on anything and everything, but we haven’t excelled at doing anything with it—until now.
Data mining allows us to finally put that data to good use by managing large datasets and analyzing them to gain new insights that we can use to better respond to threats, quickly identify new opportunities, and even gain a competitive edge. We are drowning in data but starved for insight, and this is the problem we are trying to solve with data mining.
Technology changes every day, and it becomes challenging to keep up. It’s difficult to protect ourselves from new types of fraud and other threats that technology has introduced and that are not only harder to detect but can instantly have huge impacts on an organization.
To understand how data mining can impact your organization, let’s look at an example. Imagine that you are a compliance officer and your company is under investigation.
Law enforcement claims it is involved in money laundering activities, and that one of your customers is suspected of funding illegal weapons purchases through her account.
This 72-year-old client is a retired nurse and has been a customer for 32 years—her profile does not at all match what you would expect someone accused of these crimes to look like. You have to perform due diligence and investigate the client’s activities against your different controls to see if you can find anything suspicious.
Upon investigation, you find that this client has not met thresholds, is not on any watch lists, and has not triggered any alarms. What are you missing?
Start mining your data
Data mining can help in a situation like this, but how do you get started? There are five main steps to consider when introducing data mining to an organization:
1. Identify your business objectives
What are you trying to achieve? Some common objectives include using data analytics to improve risk management strategies; identifying new customer trends and patterns to aid in customer retention; or launching a campaign to continuously clean the data circulating in the organization.
There are many different ways that data mining can help you, but not all of them have the same requirements. A clear objective will help in managing not only the resources you have, but in validating the value of the type of insights you’re looking to get.
2. Access your data sources
Data can range from large datasets and data tables stored in core business applications to a series of Excel spreadsheets that you keep on your desktop or on your department’s shared drive. You’ve identified the different sources, now if you find them you can set up different data streams.
If you know that you will consistently need this information, then you should set it up so that you have a consistent feed of that data. That way you can have the data when you need it—access is no longer an issue.
Then you need to understand the data in front of you. Typically this is where a data scientist would spend the majority of their time: assessing the state of the data, how it relates to each other, how complete is it, how do you want to treat the different values or codes that you’re seeing?
Data isn’t always clean, and it takes time and effort to understand what you’re seeing and the value you can get out of your data. If you find any gaps, this is the time to address them. You may find that you don’t actually have as much useable data as you thought.
3. Prepare the data
Once you’ve prepared the datasets, you can think about your data models. How do you decide which data model to use?
There are many to choose from and all do different things, so your choice is important because you want to use a model that best suits your business objectives. You need to be knowledgeable about the different types of models and the effects that they have.
4. Create the model
Once you’ve chosen a model, you’re ready to build something you can actually test. Over time your business needs will change, so you need to ensure that the model you build can adapt with changes not just in the business but in the types of threats facing your organization.
5. Test and deploy
Your model will need to be tested and validated—it’s a rigorous but necessary step. This should be made into a continuous process to ensure that you’re getting the best results out of your model over a longer period of time.
Stay tuned to our blog for part 2 of this webinar recap, which will offer more insight into various types of data models and how they can be leveraged to improve your compliance program.
Types of data models
There are many types of data mining techniques available today. Here are a few that can be especially useful to AML compliance programs.
The first type of data model we’re going to look at is anomaly detection. This allows you to look at any group of data, whether it be customer data, transaction data or both, and the tool will analyze the information you feed it to identify patterns. The patterns are based on all of the different data elements that you provide to the model.
The more information you feed it, the better and more precise your results will be. These patterns will show you what is normal or expected behavior based on what the data says, and will reveal outliers—anomalies—to be investigated.
This model is frequently used in fraud investigations because of its ability to quickly identify unusual behavior without requiring you to tell it the concrete definition of fraud.
Fraud is simply identified by using the patterns. If the data that you feed this model changes, so too will what it considers to be normal patterns of behavior. Now you’re looking for not just one but various types of fraud simply based on what can be considered fraudulent or non-fraudulent behavior.
Clustering is very similar to anomaly detection in terms of its ability to pattern much, but rather than singling out anomalies, it simply shows you how certain networks or groups are related based on the data you feed it—unbiased and purely data driven.
The clusters that form could represent any number of different patterns or related groups. Clustering is very effective at performing things such as crime analysis to identify patterns in where, when or how different occurrences of crimes happen.
With this information, you can make more informed decisions on how to treat different regions based on the different types of criminal activity. You could also use it to devise risk mitigation strategies.
To begin predicting where and when certain types of financial crimes can occur, we have neural nets (also known as neural networks). This model is designed to help predict decisions based on past behavior.
The machine learns from the data you feed it and improve its performance based on pattern matching.
Using machine learning, you can take what elements you already know about fraudulent activity and those customers and create a model that can actually predict those incidents before they occur.
These are particularly useful in detecting fraud early while also minimizing false-positives because instead of using thresholds and barriers to detect fraud, you’re using patterns. When a fraudster looks to change their tactics, richly trained neural nets can not only detect but adapt to the changes in behavior.
There are many benefits to data mining, including:
- Savings of time, manpower and costs
- Solving problems in real-time
- Predicting fraud
- More accurate data
- Better informed decisions
- Faster information for faster action
- Thousands of risk patterns being analyzed instantly
Applications in industry
Data mining is a powerful tool for detecting fraud, but it’s also being used in almost every industry today. Wherever there is data, there can be an application for data mining.
For example, data mining is being used in healthcare to help diagnose illness by detecting patients with similar symptoms and the outcomes of those patients. Data is now not only simply being collected, it’s being used to enrich our lives.
To learn more about data mining and how it can be used to improve your AML compliance program, contact us.