predictive-modeling-best-practices - Best practices for ecological predictive modeling
Guides predictor selection, collinearity checks, cross-validation, hyperparameter tuning, and leakage audit for ecological models.
Tags
Updated: 2026-09-15Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Define modeling objectives
- Design data splitting strategies
- Assess predictor collinearity
- Audit data leakage
- Tune model hyperparameters
- Pre-select features by importance
- Document modeling plans
Inputs
- Feature matrix (predictors)
- Target variable
- Spatial coordinates
- Candidate model list
Outputs
- cv_strategy.md
- collinearity_report.csv
- selected_predictors.txt
- tuning_results.csv
- leakage_audit.md
- modeling_plan.md
Requirements
- R environment with caret, tidymodels, blockCV, ENMeval, corrplot, usdm
- Python environment with scikit-learn, optuna, shap, scipy.spatial