“Disruptive Approaches Are Driving Change” Derek Dempsey On The Fraud Industry: Part Two

Rosalind Madge our consultant managing the role
Posting date: 2/28/2019 9:40 AM




This is the second part of our interview with Derek Dempsey, Fraud Analytics Director for FICO. To read the first part click here.


How have Data & Analytics impacted the detection, and prevention, of Fraud? 

Big Data is interesting. We want to use as much Data as possible but you have to be sure the Data you have is reliable and robust. Bad Data could lead to significant issues for anyone in Fraud detection. 

But really, Data & Analytics are the fuel that drives effective Fraud detection. Our world has used these tools very effectively for many years and we have always been at the forefront in the use of new techniques. Effective Fraud detection relies heavily on using the best Data available and the right analytical techniques to extract useful information. Furthermore, it is constantly evolving as more information becomes available. 

What impact do you think Machine Learning and AI can have on Fraud prevention and detection? 

In the last few years we’ve seen an explosion of new interest and new techniques, with lots of new players coming into the market. 

Companies like FICO have been using Machine Learning for Fraud detection for 25 years so this isn't new. What is new is a whole generation of new technologies, new companies and disruptive approaches that have been driving change. 

For us, apart from increased competitive pressures, one of the other big changes is that companies have built large Data Science teams and have a bigger appetite to build their own solutions using open source technologies. However, it’s one thing building these great models, it’s another getting them to operate effectively and correctly in the real world. The level of governance that we are subject to is enormous just to make sure that our models perform as they should. This will be a challenge to the newcomers. They are here to stay though, and should drive better performance and better Fraud detection. 

Where AI techniques are set to make an impact is in AML and compliance monitoring. These have used rule-based techniques due to regulatory pressures, but it is clear that more advanced techniques are required to provide better detection of money laundering and terrorist financing. However, businesses do require us to provide explainability but regulators are saying the will look favourably on AI usage if it can provide this. The more AI techniques are used, the more this issue of explainability is going to be important. 

Why is the development of analytics, tools and techniques so important within different industries?

I think domain knowledge of the Data and business will always be important. 'Specialisms' occur and always will. The analytic techniques, the tools and languages can be standard although some are more appropriate for different specialities. However, the differences in Data and business model always results in specialised applications to address this. 

For example, something like Card Fraud or AML require techniques that can analyse and process huge volumes of transactions, whereas something like Claims fraud or Application Fraud won't have this requirement and other factors can be more relevant. However, there is little doubt that Financial Services and Telco have actively sought AI technologies while others have been more fearful of so-called 'black box' approaches. 

How are Big Data and Data Science tools, such as Hadoop, helping combat Fraud effectively? 

There’s no question from an analytics perspective that Python and R are the two languages that Data Scientists are using. But it helps to have specialists in technical skills, such as DevOps and DataOps, to provide the technical expertise that allows them to build models and utilise Data most effectively. 

As for Hadoop, it makes sense for Data Scientists to have an understanding, but ultimately I see that skillset as one belonging to the technical specialists. This is why I firmly believe that every Data Science team should have a supporting tech function.  

What are the latest technologies helping combat Fraud? 

From an analytical perspective there has been a recent focus on graph analytics and network analytics as these can be applied to external Data. These approaches have been around for a while however and are limited to certain types of Fraud. We’re also seeing more unsupervised techniques being used as these do not rely on prior fraud case data so can be applied to a wider set of cases. 

Another new area has been adaptivity. This is a model that adapts over time depending on operational feedback from the analysts. Again this is not new but you really need to balance the impact of this new information compared to how the model is currently working and so is very challenging. As long as you can maintain a sufficient degree of explainability you can ensure the process is sufficiently well governed. 

There has been a significant move to cloud-based technologies where companies can reduce their implementation and maintenance costs. We are just about to release our new Falcon X platform, which is a cloud-based fraud platform, that will allow clients to use all the FICO capabilities but also allow them to develop their own analytics as well. 

I think potentially the biggest change will be the progressive adoption of biometric authentication. SCA is a requirement for all high-value transactions in PSD2 and this requires two-factor authentication from inherence, ownership and knowledge: so something you are, something you have or something you know. I think biometrics will start to play a big role in authentication and, hopefully, will provide much greater identity security. 

Another trend I think we may see is the growth of digital IDs. There is already a well-advanced program in the Nordics called BankID and the concept of a digital ID seems inevitable at some point. 

Derek spoke to Senior Consultant, Rosalind Madge. Get in touch with Rosalind or take a look at our latest job opportunities here.

Related blog & news

With over 10 years experience working solely in the Data & Analytics sector our consultants are able to offer detailed insights into the industry.

Visit our Blogs & News portal or check out the related posts below.

Mitigating Risk In The Financial Services Sector With Machine Learning

Data & Analytics is an industry that is constantly evolving and is always using the latest technology to innovate its services and capabilities. More recently, these advancements have moved in areas such as Artificial Intelligence (AI) and Machine Learning (ML). Machine Learning is a method of data analysis, under the branch of Artificial Intelligence, that allows systems to learn from data, identify patterns and ultimately make decisions with little to no human intervention. Used across a vast range of sectors, this arm of Data & Analytics has become widely popular, especially within highly-advanced industries such as Finance.  Since the 2008 financial crash, at the top of the agenda for many Financial Institutions (FIs) was, and still is, the need to protect business, increase profitability and, possibly most importantly, address the abundant inadequacies of risk management. This includes risks posed by consumers such as liquidity, insolvency, model and sovereign, as well as any internal process and operational risk the FIs may also be facing through any failures or glitches.   Machine Learning has played a crucial role in improving the quality and precision of FIs risk management abilities. In HPC Wire, it has been reported that the use of AI and ML within the financial sphere to mitigate risk, improve insights and develop new offerings may generate more than $250 billion for the banking industry.  How does ML work? By using incredibly large data sets, drawn (with consent) from consumers, ML can learn, and predict, patterns in consumer behaviour. This can be done in one of two ways: through supervised learning tools, or unsupervised learning tools.  Explained by Aziz and Dowling; “In supervised learning you have input data that you wish to test to determine an output. In unsupervised learning, you only have input data and wish to learn more about the structure of the data.” How do banks use ML to mitigate risk? In FIs, a mix of the two ML tools are used. Most commonly, we can expect to see learning systems such as data mining, neural networks and business rules management systems in play across a lot of banks. These models work in tandem to identify relationships between the data given from the FIs and their consumers – from their profiles to their spending habits, credit card applications to recorded phone calls – which then build ‘character profiles’ of each individual customer. The process can then begin, spotting signs of potential risk factors. This may include debt, fraud and/or money laundering.  Here we break down two key examples.  Fraud Thanks to ML, customers have become accustomed to incredibly quick and effective notification of fraudulent activity from their banks. This ability from FIs comes from large and historical datasets of credit card transactions and machines which have been algorithmically trained to understand and spot problematic activity. As stated by Bart van Liebergen; “The historical transaction datasets showcase a wide variety of pre-determined features of fraud, which distinguish normal card usage from fraudulent card usage, ranging from features from transactions, the card holder, or from transaction history.”  For example, if your usual ‘character profile’ is known by ML tools to spend between £500 - £1000 per month on your credit card, and suddenly this limit is overtly exceeded, fraudulent activity tags will be alerted, and the freezing of your account can be done in real-time.  Credit applications When borrowing from a bank or any other FI, consumers must undertake a credit risk assessment to ensure that they have a record of paying back debt on time, and therefore not adding greater cost, and risk, to the lender.  Traditionally, FIs have approached credit risk with linear, logit and probit regressions but, serious flaws were found in these methods, with many applications being left incomplete. In this space, the evidence for the effectiveness of ML is overwhelming. Khandani et al. found that FIs using ML to analyse and review credit risk can lead to a 25 per cent cost saving for the FI involved.   These ML models come in various shapes and sizes, with the most common being instantaneous apps or websites which allow users and their banks to have access to real-time scoring, data visualisation tools and business intelligence tools.  The risk of risk management with ML Like with any AI or ML application or tool, there will always be cause for concern and real need to always remain vigilant. While ML has shown to be an invaluable tool across lower risk areas, the complexity of more statistical areas of banking, such as loans, has proven to be an Achilles heel for the technology. This usually stems from bias, a perpetual problem for AI and ML across all industries.  Technology Review notes that “There are two main ways that bias shows up in training data: either the data you collect is unrepresentative of reality, or it reflects existing prejudices.”  Data, analytics, AI and ML are notoriously non-diverse working sectors. The person behind the screen creating learning algorithms tends to be white and male, and very unrepresentative of the whole society that the machine learning tool will serve. Over the years, we have heard numerous accounts of unfair and unjust machines, which have learned from a very narrow and unrepresentative dataset, which stems from the lack of diversity amongst the employees within the Data & Analytics industry. For example, Microsoft’s racist bot and Amazon’s sexist recruitment tool, both clear examples that ML and AI are not ready to be used on their own, and humans still need to play an integral part in decision making.  Banks and FIs must be aware of the, potentially lethal, consequences that bias in ML may present. Lenders must be careful to ensure they are working within the guidelines of fair lending laws and that no one group of people are being penalised for no reason other than issues within the technology and its algorithms. It is vital that the humans behind the technology don’t rely on ML to provide them with an answer 100 per cent of the time but, instead, use it to aid them in their decision making when it comes to risk mitigation.  If you’re looking for a role in Data & Analytics or are interested in finance or Risk Analytics, we may have a role for you. Take a look at our latest opportunities or get in touch with one of our expert consultants to learn more.  

Risk Analytics & Your Job Search

Risk Analysis is a daily part of our decision-making process. Its influences are felt in how we prioritise projects. Hello, project management. It determines when we take breaks. And in the working from home or remote working culture, lines between personal and professional life are soon blurred. Our decision-making ratchets through microanalyses of next steps, words to use, when to log off the computer, and what time to make lunch or dinner. Almost without realising it, we're using the SWOT matrix determine risk. That's just in our personal day-to-day lives. Organisations have been using these strategic planning techniques as a daily part of their enterprise lives. So, imagine this. You determine your strengths, weaknesses, opportunities, and threats (SWOT) to write your CV, land the interview, and get the job. Aren't you already on your way to helping your company determine their risk on a bigger scale?  Isn't Risk Analysis the Same as SWOT? Not exactly. While you are identifying, measuring, and analysing issues, the idea is to avoid or lessen risk. In the movie War Games, the main character decides he's going to play a game. But, neither the computer nor the main character understands the game is not a game. They have no idea what risk they've unleashed as keystrokes begin to lead to war.  Militaries use Risk Analysis regularly to determine if war should begin, estimated casualties and cost just for a start. But, in the end, it's usually decided peace keeps the world away from war after all the risk has been analysed. Though most businesses and individuals don't have the risk of war at their doorstep, there are extenuating circumstances which must be determined to avoid risk. This can be anything from natural disasters to legislation to physical requirements and locations. The SWOT technique is a tool within strategic planning as are Risk Analytics. Let's take a look at the benefits of Risk Analysis.  It can help your business improve security, manage costs, and plan for any surprises. Whether it helps you in your job search or manage your business once you're hired, a combination of SWOT and Risk Analytics can help your decision-making process shine. Done well, Risk Analysis is an important tool for managing costs associated with risks, as well as for aiding an organisation's decision-making process. SWOT your way to a Successful Job Search So, first, let's take a look at what SWOT is and isn't. This type of analysis is a tool most often found in strategic planning for organisations. But, it's not the only tool when assessing risk. Like any data profession, you'll want to gather, collect, and analyse the information before you. And when it comes to the job search, knowing yourself and your self-awareness levels may play a bigger part than you imagine. Skipping ahead to the interview from your 'foot in the door' CV and cover letter both delivered via video, of course, you begin to plan for your interview. "What are your biggest strengths and weaknesses?" your hiring manager asks. You've come well-prepared for this question because you've done your SWOT analysis. From the moment you decided your areas of expertise, your roles in organisations, and any areas you know you need to improve but can turn into a positive. You've pre-assessed your strengths, weaknesses, opportunities or obstacles, and threats. Strengths – What characteristics do you possess which gives you an advantage over your competition? Have you cross-trained across a variety of departments? Do you have a knack for telling a compelling data narrative story to help make leadership make an informed decision? Weaknesses – Where could you use more training? Are you looking for a business that offers it in-house or do you need a certification or class to give you a leg up over your competition? What puts you at a disadvantage and how would find a work around? What can you do to improve? Opportunities or Obstacles – Has your experience taught you a new way of doing things within the industry? Does your experience extend from working alone to strong member of a team? Was the team in-house or scattered around the world? How did it affect your working style and what did you learn from it? Threats – Threats may seem at odds with a job search, but…that was then. This is now. Threats are simply external forces which cause trouble in your planning. In other words, this is your risk assessment. What risks are involved and of those risks which will have less effect and which will have more on your desired outcome? Performing this technique in your personal and professional life helps you peel back the layers of you. What has your education prepared you for? Your work experience? The projects you've chosen or been assigned to? It's all leading somewhere, right? This is where you match your strengths to opportunities. Hello, dream job. If you're looking for your next role in the Data & Analytics, we may have a job for you. Take a look at our current opportunities or get in touch with one of our expert consultants to learn more.  

RELATED Jobs

recently viewed jobs