CLASS 3 - Course 3: Building Gradient Boosting Models

Abstract: This course focuses on how to approach model building when using modern machine learning models in general, and specifically gradient boosting.  Guidelines that made sense when building a linear regression model on 200 data points are no longer applicable when building a gradient boosting model on 20,000 data points.  We will present an approach to model building in which the data scientist builds understanding of the data, interacts with the model, and can apply domain expertise to make improvements.

  – Module: Model Building Principles

  – Module: The train/test paradigm

  – Module: Iterating and Improving your Model

  – Module: Early Stopping / Hyperparameter Optimization

  – Module: Failure Analysis and Feature Engineering

  – Module: Evaluating and Comparing Models