How to Deal With Concept Drift in Production with MLOps Automation

Iguazio’s Data Scientist discusses how to detect and handle problems that arise when models lose their accuracy and how to implement concept drift detection and remediation in production. He shows how to automate MLOps processes at scale, to handle drift detection using open source tools.

In this session you will learn:

• Why and how models lose their accuracy due to concept drift and the problems this poses for data scientists and ML engineers
• What methodologies exist for concept drift detection and the different options available for handling concept drift once it has been detected
• How to implement concept drift detection and remediation in production, and:
◦ Automatically detect concept drift by monitoring and understanding whether your models have been impacted
◦ Harness automated tools for adjusting models
◦ Utilize online models that adapt to shifting data
◦ Monitor concept drift in an ongoing manner by setting up alerts and tracking error rates
• How to automate MLOps processes at scale to handle drift detection using open source tools – a live demo will be shown