ML Model Changelog

Every StoQuant production model retrain is logged here with its validation AUC, sample size, champion/challenger status, and top-5 feature importances. Updated daily as new retrains finish.

What each entry contains

• Timestamp — UTC training completion.
• Model family — LightGBM, CatBoost, XGBoost, or stacked meta-learner.
• Validation AUC — out-of-fold area under the ROC curve.
• Sample size — labeled (ticker, date) rows used for training.
• Champion flag — whether this model was promoted to serve predictions.
• Top-5 features — highest-gain splits by permutation importance.

Why publish this

Institutional-grade transparency. Retail investors deserve to see when a model's AUC degrades or when a feature importance shifts — the same way a quant fund's risk committee would.