Fraud detection

Think

Smart detection

Although better than a random choice, the process in place used sophisticated but static business rules, thus limiting the ability to distinguish new fraud patterns. We have innovated by proposing to include advanced learning algorithms in the process of detecting fraud on treatment files.

Make

Advanced predictive models

The success story begins with a “Proof of Value” conducted by Eulidia and its partners: Dataiku & Teralab, providers of a sovereign environment for storing sensitive data and a studio, a productivity factor for our teams in developing the model. In a few weeks, our client had access to a descriptive analysis allowing a better understanding of the variables that explain fraud in order to manage its network of partners, as well as a risk model for scoring new treatment files.

Scale

Customer managed process

The infrastructure and the model are now deployed in production. The internal team was trained to update the model via the DSS Dataiku studio as well as to supervise the predictive process to allow auditors to prioritize their controls and significantly reduce the impact of this phenomenon.