GlueRatGlobal Robotics · classification laboratory

Follow the rules.

A decision tree learns a sequence of yes/no questions. Change how large the tree may grow, then change one passenger and watch the route—and prediction—update immediately.

PYTHON206 · Titanic dataset
1 · grow the model
3
8
Model report card

A compact tree with generalisable rules

Training score and held-out score should be read together. A widening gap is evidence that the tree may be memorising quirks.

Training accuracy—
Held-out accuracy—
Generalisation gap—
Leaves—
The passenger’s decision route
each card is one learned if/else rule