Classification with Decision Trees

Overview

This week introduces three essential machine learning concepts:

• Decision Tree: A classification procedure as an alternative to logistic regression.
• Hyper-Parameter Tuning: Evaluate different forms of a model set in advance of model estimation, such as the number of features, with a grid search.
• Move beyond discovering the best model for a given estimation procedure, also exploring different algorithms and models for the most accurate prediction.

The machine learning analyst leverages massive computer power to check almost every alternative on the testing data in pursuit of the best model in terms of fit metrics. The better the fit to the testing data, the better the prediction.

Concepts: Classification with Decision Trees

Decision Tree Concepts

Decision Trees Concepts [43:02]

Access each chapter from the three horizontal lines menu in the top-left of the video. Click again on the lines to remove the menu.

The last two chapters, Hyperparameter Tuning and Random Forest (Ensemble Methods), were re-recorded to exactly match the more extensive revisions in the written text. The preceding chapters were revised less in the written material and so were not re-recorded.

• 00:14 Overview
• 01:05 Split #1
• 11:36 Split #2
• 18:06 Split #3
• 24:15 Overfitting
• 29:20 Hyperparameter Tuning
• 36:30 Random Forest
• 42:50 End

Decision Trees: sklearn

Decision Tree Template [pdf template]

Decision Tree Template [29:52]

Homework

Short-Answer Problems

Analysis Problems