Deep Learning: Autoencoder for Model Monitoring

The previous white paper in this series, Open Banking: Multi- Layered Self-Calibrating (MLSC) models, discussed the use of self-calibrating/semi-supervised machine learning (ML) models for open banking. Regardless of the ML approach, monitoring the operational behavior of models can be a very challenging task. Off-board model performance data and model performance monitoring is a crucial part of a model governance process.

This white paper focuses on autoencoder technology and its use in monitoring how the data and derived features in the production environment are changing in comparison to the data on which models were developed.