Reference · Glossary

Decision tree

Last updated

A **decision tree** splits data with if/else rules to predict a class or value. Easy to visualize; easy to overfit if grown too deep.

#When to use

Tabular problems where interpretability matters: credit rules sketches, triage policies, feature interaction exploration.

#When not to

Raw text/image tasks (use deep models), or when a single deep tree memorizes noise — prefer forests or regularized models.

#Quality checklist

  • Limit depth / min samples per leaf
  • Compare against a simple baseline
  • Inspect feature importances with skepticism
  • Validate on held-out time periods if data drifts

#Learn next