Detecting dataset and prediction shifts for machine learning models

Digital learning, Lesson

20 minutes

2 hours

Beginner

Professionals, Students

This nugget will guide the learner through different metrics for the detection of dataset shifts when monitoring machine learning models.

We will build a basic toolbox for machine learning monitoring with unsupervised and supervised solutions.

This guide will rely on bilinear models for regression problems.

Learning outcomes

The learner will be able to define metrics for dataset shifts and types of dataset shifts with supervised and unsupervised approaches
The learner will be able to interpret the diagnosis metrics and map them to the possible causes of data shifts
The learner will be able to differentiate between supervised and unsupervised monitoring metrics for causes and types of model performance degradation