Solenix

The European Central Bank (ECB) periodically collects data from European banks for supervisory purposes. To complement the ECB’s quality assurance and assessment tasks, Solenix has participated in a project with the aim of improving the plausibility checks performed by the ECB with Machine Learning (ML).

During the project a prototype was developed. With this prototype the ECB can:

• Automatically discover new plausibility checks. These new checks complement the validation rules already in use by the ECB.
• Automatically discover potential anomalies in the data that might have gone previously overlooked.

The ML prototype uses Explainable Artificial Intelligence (XAI) approaches to provide domain experts with explanations on the newly found plausibility checks. These explanations increase confidence in the anomaly detection results as they match domain experts understanding. In addition, the XAI explanations allow domain experts to include the new plausibility rules found by ML in the ECB validation system.

The ECB and Solenix have written the paper “Discovering new plausibility checks for supervisory data” describing this work. The paper has been published in the ECB Statistics Paper Series.