Regulatory efforts may prove valuable, but banks and law enforcement still need to use SARs more effectively, and that’s where cognitive analytics for AML can be a tremendous boon.
Behavioral profiling approaches are extremely important in tackling fraud that happens when banks share financial data with third parties through application programming interfaces.
Uncovering hidden patterns in money movement makes machine learning for AML a very attractive enhancement to existing AML operations.
The mathematical optimization community already has a clear set of standards for MIP benchmarking - these should be followed.
In my role at FICO I have the privilege of engaging in strategic planning discussions with senior banking executives across the globe. Most of these meetings as of late are similar: We are improving the customer on-boarding user experience, both through improved user interfaces, but also through improved analytics to bring everything we know about the customer to bear in our interaction…regardless of product line or channel. Right now we are working on our data lake. At which point, I nod. Then I ask: OK, so who is in charge of the customer journey that you are transforming? At a strategic level, banks understand the challenge they are facing and where their organizations need to go in order to thrive. At a tactical level, the decisions become much more complicated. Who owns the customer? Is that new data source worth the risk of potentially slowing down my origination process? How seriously... [Read More]
My presentation on “Transparency by Design: An Explainable AI Approach to Credit Risk Scoring” discusses our application of recent machine learning technologies to challenge the US FICO® Score.
“There are three kinds of lies: lies, damned lies, and statistics” – Attributed to Benjamin Disraeli, British PM (1874-1880) In assessing the predictive value of new scores or new data, significant weight is given to model performance measures. Examples of these measures — which quantify the effectiveness of a predictive model in distinguishing between the outcomes being predicted — include the Gini coefficient, KS (Kolmogorov-Smirnoff statistic), and ROC Area. These measures are appealing because they allow a straightforward means of comparing the relative effectiveness of two or more models. Think of them as scoring the scores — the score or data with the highest “score” looks like the best one to buy or use. Score and data vendors often structure their validations in a way that produces a more compelling case for their product. And sometimes their efforts fall into the grey area of gaming — producing misleading results by tweaking... [Read More]
More and more, businesses realize that treating every customer interaction as “one size fits all” is a recipe for irritating them, losing their attention, and ultimately having a competitor take them away. Conversely, consumers are more likely to be loyal to those brands that demonstrate empathy and understanding of what matters to the customer. When FICO introduced “Decision Management” as an integrated technology and analytics platform offering several years ago, it empowered organizations to codify their interaction logic, empower it with predictive and prescriptive analytics, and rapidly build and deploy scalable, 360-degree customer engagement solutions. A platform approach allowed them to connect and ultimately centralize decisions across business units and functions, powered by business users who could test and deploy scenarios and strategies, without needing IT or other specialized assistance. As the centralized model gains traction throughout the entire business, customer engagement is powered by a “sense and respond” model... [Read More]
Fraud analytics is an umbrella term covering a lot of technologies — let’s look at the two big categories.
Because corporate credit cards have different business-oriented spending patterns than consumer credit cards, we tackled the corporate card fraud detection problem with FICO’s collaborative profiling technology.