LightGBM (cross-model enriched)
Fraud Detection.
Supervised fraud classifier (LightGBM, isotonic-calibrated). PR-trio (2026-05-02) un-deprecated the PR #14 stub by adding the missing trainer + consumer.
Foundation · LightGBM (cross-model enriched)
The honest description.
Supervised fraud classifier (LightGBM, isotonic-calibrated). PR-trio (2026-05-02) un-deprecated the PR #14 stub by adding the missing trainer + consumer.
How it works.
Transaction context (amount, time_of_day, merchant_category_score, payment_method_score), user history (prior_fraud_count, days_since_account_open, is_new_payee), velocity (1h, 24h). Cross-model: cm_p2p_is_p2m, cm_anomaly_score, optional cm_forgery_score. P(fraud) ∈ [0, 1]. A rule-based heuristic score currently serves; the supervised LightGBM classifier (isotonic-calibrated, distinct from unsupervised anomaly_detector) is the intended target, not yet served — models:/fraud_detection/Production is absent by default.
What it learned from, and how we watch it.
Inputs: Transaction context (amount, time_of_day, merchant_category_score, payment_method_score), user history (prior_fraud_count, days_since_account_open, is_new_payee), velocity (1h, 24h). Cross-model: cm_p2p_is_p2m, cm_anomaly_score, optional cm_forgery_score.
Drift baseline not yet documented for this model.
- PR-trio (2026-05-02): supervised counterpart to anomaly_detector.
- Falls back to rule-based heuristic when MLflow model not loaded.
Across the house and the API.
No public routes documented yet.
Every release, traceable.
Data Mukodi holds about you is subject to DPDP §12 erasure (use /consent in the app), and any concern about this model can be raised via [email protected] under DPDP §13 grievance, or via /security.