From the course: MLOps Essentials: Model Deployment and Monitoring
Join today to access over 24,500 courses taught by industry experts.
Feature drift basics
From the course: MLOps Essentials: Model Deployment and Monitoring
Feature drift basics
- [Narrator] Having discussed concept drift, let's now focus on the other type of drift called feature drift. A feature drift occurs when the distribution of values or classes of features change over time. This indicates a change in the features that are different from those that were used during model training. Looking at the conditional probability formula, there is a change in the probability of X for feature drift. One important question is whether the occurrence of a feature drift would mean that a concept drift is also happening. There may be a concept drift, or there may not be. A change in probability of X may be accompanied by an equivalent change in the joint probability of X and Y, leaving probability of Y given X unchanged. Let's look at some examples of feature drift. In the same example of predicting whether a patient is diabetic, let's look at the distribution of classes over a few months. In the first three…
Practice while you learn with exercise files
Download the files the instructor uses to teach the course. Follow along and learn by watching, listening and practicing.
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.
