Feature monitoring
We have discussed how important feature monitoring is in an ML system a few times in the book. We have also talked about how a feature store standardizes feature monitoring. In this section, let's look at an example of feature monitoring that can be useful for any model. As feature monitoring is calculating a set of statistics on feature data and notifying the data scientist or data engineer of changes, it needs the latest features used by the model.
In this section, let's calculate the summary stats on the feature data and also feature correlation, which can be run on a schedule and sent to people of interest regularly so that they can take action based on it. As mentioned in the last note of the Model training section, the steps to fetch the features are the same as what was done in that section. Once you have all the features, the next step is to calculate the required stats.
Important Note
Please note you may have to install additional libraries...