Regulatory Anomaly Detection

Data Science

An advanced analytics group charged with detecting anomalous transactions across the regulatory space.

"Using Machine Learning and Automation Algorithms to Drive Better Insights."

Business Challenge:

  • Detect both simple and complex anomalies across the regulatory space.
  • Provide transparency into why a transaction is deemed anomalous in a way business users can understand and respond to.

Technical Solution

  • Data feeds come from several sources at various reporting times and with proprietary schemas.
  • There is no ‘known anomaly’ data set to train from, yet models must provide a justifiable confidence metric.
  • Models must provide transparency into criteria used to identify given anomalies.

Results Achieved

    Atyeti was able to structure an advanced data engineering and analytics team which validated and homogenized the data. By constructing a multi-level analytics pipeline, paired with a complex ensemble of deep learning algorithms, truth determination and explainer networks, we were able to meet all business requirements and technical objectives.