xgb-modeling

To train an XGBoost binary classification model with AUC and KS evaluation, use this Skill to process a 0/1 target column and structured inputs to generate decision-grade metrics, stability diagnostics, and a detailed markdown report.

PSI stability analysis in financial binary classification modeling measures feature and model score distribution shifts across time windows, validating out-of-time performance to ensure reliable credit risk predictions.

You can compare multiple feature selection schemes for XGBoost in one run by passing complex feature sets via a config file, generating side-by-side performance comparisons using IV, PSI, and correlation selectors.

XGBoost works with scikit-learn and optbinning for feature engineering by integrating structured data inputs and automated binning selectors to standardize feature schemes and produce portable model artifacts.

The best way to evaluate XGBoost models for finance-style risk scenarios is to calculate AUC, KS, Gini, Lift, and Bad Capture Rate metrics, alongside time-window performance and PSI stability analysis.

Your XGBoost model output requires a ModelCard and result.json manifest to provide portable, no-platform-coupling workflow artifacts that document training logic and standardize downstream skill consumption.