Predictive analytics and machine learning hold the potential to improve outcomes and enhance decision making across a variety of data-rich industries and processes. The same applies to both financial fraud prevention and healthcare challenges. Although the terminology may differ, the methodologies underpinning these capabilities are consistent whether applied to these or other verticals.
We know that fraudsters continually evolve their tactics to attack the areas of greatest vulnerability, and have access to ever more powerful tools to support their schemes. While not as nefarious, the spiraling cost of healthcare and medical needs of an aging population increases the urgency to maximize efficiencies in healthcare as well.
With these challenges in mind, let’s consider how machine learning and other tools can help professionals in the financial services, healthcare and other fields yield results beyond those possible with traditional labor-intensive, rules-based approaches.
Seeing the Random Forest through the Decision Trees
Decision trees have been employed in risk (and other) analyses for generations, long before big data and machine learning had entered the lexicon. When performed manually, this process is essentially limited to a series of “forks in the road” at which a series of two or more logical outcomes are considered. As computing power grew, analysts were empowered to consider an ever-increasing number of decision points and alternatives. Gradually such decision points become essentially continuous, driven by a set of more complex algorithms.
ARGO’s white papers covering fraud prevention and healthcare alike outline several such machine learning constructs that can be used to create probabilistic models. These algorithms carry arcane names like Fellegi-Sunter Probabilistic Match and Naïve Bayseian, but the underlying concepts are fairly intuitive. Random Forest Classifier, for instance, can correct for bias when a model reacts to noise rather than the true underlying relationship between variables. Particle Swarm Optimization “supervises” machine learning by considering the position and velocity of various traits within a data set.
These techniques help to answer three key questions: what happened; why did it happen; and what is likely to happen in the future? A key benefit is that the process can detect low-confidence patterns that escape the naked eye, allowing these factors to be integrated into models and accelerate continuous business improvement - which after all is the core objective.
Deterring Fraud, Preventing Relapses
With predictive analytics, ARGO was able to help a large U.S. financial institution dramatically reduce the number of false positives and better identify suspect first-party and third-party fraud transactions, focusing resources where they can actually contribute value. This delivers multiple benefits: reduced fraud losses, better customer experience and more efficient internal processes.
In the healthcare realm, ARGO has deployed predictive analytics to predict client attrition, identified duplicate records and detected the likelihood a patient faces elevated readmission risk so that preemptive measures can be taken. As with the fraud prevention example, the ultimate outcome is lower cost and a better customer experience.
The use cases and jargon may vary by field, but the potential for predictive analytics in banking and healthcare to deliver value across a wide array of industries is clear. To learn more, check out our white paper below.