Task: Binary classification — predict viral (1) vs not (0), where the label reflects YouTube engagement (definition as in your project; training excludes direct YouTube count/rate features from X to reduce leakage).
Combined XGBoost (viral)
Tuned XGBClassifier; CV scoring ROC-AUC. Same feature matrix convention as RF. Feeds the stacking layer via predict_proba[:, 1].
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support