ML model continuous monitoring
Once you’ve deployed your model into production, you need to monitor model performance continuously to ensure that it is performing as expected. We recommend using Vertex AI, which provides two ways to monitor your ML models:
- Skew detection, which looks for the degree of distortion between your model training and production data.
- Drift detection, which looks for drift in your production data. Drift occurs when the statistical properties of the inputs and the target change over time and cause predictions to become less accurate as time passes.
For skew and drift detection, we recommend setting up a model monitoring job by providing a pointer to the training data that you used to train your model, and then tuning the thresholds that are used for alerting to measure skew or drift occurring in your data.
You can also use feature attributions in Vertex Explainable AI to detect data drift or skew as an early indicator that model...
