liveBook · Manning

Isotonic regression

As explained in Machine Learning System Design

A regression technique that fits a monotonic function to data while preserving the specified ordering relationship.

Isotonic regression serves as a fast, simple postprocessing model for calibrating forecasts when elasticity assumptions create prediction problems. section 9.4.1

Choose between two improvement paths:

  • Postprocess forecasts with isotonic regression when a quick, ad hoc correction is appropriate. section 9.4.1
  • Improve the underlying model through modeling, feature engineering, and preprocessing when a slower, more generic solution is justified. section 9.4.1

How Machine Learning System Design covers it

How deep it goes
Positions isotonic regression as an ad hoc calibration option without detailing its fitting procedure or implementation.
Examples
Uses isotonic regression as an additional model for postprocessing forecasts affected by elasticity-capture problems.
What it assumes
Assumes familiarity with forecasts and elasticity assumptions, and does not teach either from zero.
What you can do after
Apply it via prediction postprocessing to calibrate forecasts when elasticity-related problems appear.
How it connects
Connects isotonic regression to forecast calibration, elasticity assumptions, residual analysis, and model improvement.
Cover of Machine Learning System Design
liveBookmanning.com
Read Machine Learning System Design on liveBook Get the book at manning.com

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Machine Learning System Design. The section links point to the parts of the book it draws on.