The EU Market Abuse Regulation (MAR) requires institutions to monitor transactions and develop specific algorithms to check for possible abuse covering insider dealing, market manipulation and other categories. One of the challenges is that calibrated monitoring thresholds tend to be conservative and consquently produce a high number of false positives. These must then be manually investigated by compliance departments.
A very interesting post on the ION Markets Blog, written by our colleagues Davide Bonamico and Serena Manti, explains how Machine Learning (ML) can be applied to this problem and shows very promising results.
Please read at How ML can improve alarms classification to detect market abuse.
