Cortex AI logo placeholderCortexLabAI · learning · engineering
Model Atlas · Tree ensembles

Random Forest

An ensemble of decorrelated decision trees that captures nonlinear interactions with limited preprocessing.

core mathematical viewf(x)=1/T Σₜ fₜ(x)
Mental model

Understand it before memorizing it.

Many trees vote so individual tree variance is averaged away.

Best fit

Where this model earns its place

Tabular classification

Start with a simpler baseline then compare this model using the same evaluation split and operational constraints.

Tabular regression

Start with a simpler baseline then compare this model using the same evaluation split and operational constraints.

Strong nonlinear baseline

Start with a simpler baseline then compare this model using the same evaluation split and operational constraints.

Strengths and limits

Trade-offs matter more than popularity.

Strengths

✓ Robust

✓ Handles mixed feature scales

✓ Feature importance options

Limitations

△ Large models

△ Probability calibration can be weak

Evaluation

Metrics to watch

PR-AUCInterpret with the product objective and error cost.
ROC-AUCInterpret with the product objective and error cost.
RMSEInterpret with the product objective and error cost.
LatencyInterpret with the product objective and error cost.
Production checklist

Before it reaches real users

  1. 01

    Control tree count/depth

    Document the assumption and instrument the condition so regressions can be detected.

  2. 02

    Measure inference cost

    Document the assumption and instrument the condition so regressions can be detected.

  3. 03

    Validate feature importance carefully

    Document the assumption and instrument the condition so regressions can be detected.