ML Model Monitoring Best Practices

Thursday, November 15th, 2018

Data scientists often spend a significant amount of time building and refining sophisticated predictive models, but this is just the beginning. In order to know if a machine learning model is useful, you need to see how it behaves in the real world. The ultimate goal of a predictive model is providing business value, but this requires answering several questions:

  • How does my model evolve with time?
  • Can I take action on this model?
  • How does my current model compare with a past version? Should I revert?  

Model monitoring is the key to understanding model performance and quantifying successful model deployment, but it proves to be extremely challenging for most organizations to do successfully.

Join us for a 30 minute webinar on best practices in model monitoring. Dr. Miriam Friedel will walk through why model monitoring is essential, discuss its critical components of successful monitoring, provide guidelines on determining the right metric to track, and give a demonstration of what successful model monitoring looks like on Skafos.