A global finance firm that helps leading organizations advance in three key areas: disputes and investigations, corporate finance, and performance improvement and advisory was tasked with building a state-of-the-art machine learning (ML) model that could detect days of anomalies behavior within market trading behavior on behalf of a large-scale investigation.Â
phData created an anomaly detection model on Azure ML and leveraged MLFlow as a way to track experimentation
As a part of a large-scale investigation, a global finance firm needed to apply data science to assist in detecting instances of financial anomalies. The volume of underlying data to be analyzed was large, but the number of days of known anomalies was rather small.Â
The client knew they needed to partner with a team of data science experts, as anomaly detection models of this caliber can be difficult to create due to the changing behavior that drives suspicious behavior and the small number of known cases. phData was brought in to assist and they proposed building a state-of-the-art solution that involved an anomaly detection model on Azure ML and leveraged MLFlow as a way to track experimentation.
phData was chosen for this engagement because of our deep expertise in machine learning and our ability to take known approaches and modify them to address novel problems.
phData was able to build an accurate and robust anomaly detection model to determine unusual behavior in over 10 markets for nearly a decade’s worth of data. Through the use of this model, the global finance firm was able to provide a list of instances of suspicious behaviors occurring in financial markets. This information could result in enormous amounts of savings from shutting down suspicious accounts.
Need assistance developing your company’s Machine Learning strategy?