Monitoring your models
The ML lifecycle does not end at deployment. Once a model is in production, we want to monitor the input data and output results of the model. In Chapter 4, we explored two key features of Databricks Lakehouse Monitoring integrated with Unity Catalog: Snapshot and TimeSeries profiles. Snapshot profiles are designed to provide an overview of a dataset at a specific point in time, capturing its current state. This is particularly useful for identifying immediate data quality issues or changes. On the other hand, TimeSeries profiles focus on how data evolves over time, making them ideal for tracking trends, patterns, and gradual changes in data distributions.
Expanding on these capabilities, Databricks also provides an Inference profile, tailored for monitoring machine learning models in production. This advanced profile builds upon the concept of TimeSeries profiles, adding critical functionalities for comprehensive model performance evaluation. It includes...
