Skill: predictive-modeling-best-practices

Domain: CV · Tuning · Leakage · Collinearity · Overfitting
Phase: 1 — Foundation
Used by: run-sdm-study, build-fire-risk-map

Purpose

Ensures that any predictive model in the project is built with sound ML practices: proper data splitting, cross-validation strategy, hyperparameter tuning, collinearity reduction, leakage prevention, and overfitting diagnosis.

When to Invoke

• Before fitting any algorithmic model (MaxEnt, BRT, Random Forest, ANN, GLM for prediction)
• When designing the validation strategy for a modeling study
• When the user asks about feature selection, predictor reduction, or model tuning