Mapfre trains an AI with artificial data to detect fraud in Home claims

MADRID, 22 Abr.

Mapfre trains an AI with artificial data to detect fraud in Home claims


Mapfre is working on an experimental project in which Artificial Intelligence is trained with artificial data, also generated through AI, with the aim of detecting fraud in Home claims.

According to the insurer, it has created an AI system that applies machine learning and graph analysis by examining multiple historical data points to detect fraud patterns. In this way, when a claim is registered, the system applies a note to it. If this note indicates the possibility of fraud, it is referred to the claims team, which, together with the investigation team, carries out the necessary investigations to decide whether there is fraud or not.

Mapfre indicates that this process has helped its claims team improve efficiency and accuracy in detecting fraud and has led to cost savings. This AI was first applied to the Car business, and later, given the positive results, it was extended to the Home business.

The obstacle that Mapfre encountered is that AI detection models use historical data, but in Hogar there was a greater imbalance between fraudulent and non-fraudulent claims. To solve this and better train the AI, the insurer has decided to generate "synthetic data" (artificial, not real), to feed the algorithm.

Thus, Mapfre has decided to implement a generative AI model called CTGAN (Conditional Tabular Generative Adversarial Networks). This model generates synthetic tabular data and preserves privacy.

"This strategy allows us to overcome the imbalance and scarcity of historical fraudulent claims, improving the algorithm's ability to identify fraud patterns in a more precise way. By generating a more balanced data set, the company achieves that its fraud detection models in Home insurance policies are much more precise," the insurer argues.