Reading Surrogate Decision Trees in Driverless AI

This lesson explains how to read the surrogate decision tree in Driverless AI's interpretability report, which trains a simple decision tree to approximate a more complex model's predictions. Using a credit score dataset, it walks through how variables like outstanding debt, interest rate, and credit mix sit higher or lower in the tree based on their importance, and how thick branches mark common decision paths while thin branches mark rare ones. It also shows how to search for an individual application by row ID to see which branch and leaf it falls into, along with the average prediction and the underlying pseudo code and Python code.