Macroeconomic Recession-Signal Modeling Pipeline
A straightforward pipeline that prepares monthly U.S. macroeconomic data (unemployment, yield curve, industrial production, inflation, and policy rates), creates year-over-year and lagged features, and trains three supervised models — logistic regression, random forest, and XGBoost — to classify recession vs non-recession months.
TThe workflow includes data cleaning, resampling, feature engineering, time-based train/test splitting, ROC-AUC evaluation, and model interpretation using PCA and SHAP.