fraud-detection / models /autoencoder_meta.json
fikri0o0's picture
2026-06-05: deploy fraud detection dashboard (LightGBM + GNN + autoencoder, SHAP, drift, live scoring)
99bc19c verified
raw
history blame contribute delete
497 Bytes
{
"model": "Autoencoder (unsupervised, legit-only training)",
"trained_at": "2026-06-05 21:44",
"test_metrics": {
"pr_auc": 0.13483,
"roc_auc": 0.86586,
"f1_at_best": 0.05365,
"best_threshold": 0.46032,
"precision_at_best": 0.23381,
"recall_at_best": 0.0303,
"precision_at_100": 0.09,
"recall_at_1pct": 0.43636,
"total_cost": 16556.00456,
"cost_at_half": 16755.96909,
"n": 555719,
"n_fraud": 2145
},
"n_features_in": 88
}