Handling changes to the feature set during development
Model development is an evolving process. So are models – they evolve over time. Today, we may be using a few features for a specific model, but as and when we discover and try out new features, if the performance is better than the current model, we might end up including the new features in the model training and scoring. Hence, the feature set may change over time. What that means with the feature store is some of the steps we performed in Chapter 4, Adding Feature Store to ML Models, might need to be revisited. Let's look at what those steps are.
Important Note
The assumption here is feature definitions change during model development, not after production. We will look at how to handle changes to the feature set after the model goes into production in later chapters.
Step 1 – Change feature definitions
If the features or entities change during the model development, the first step is to update...