elevator-maintenance / metrics.json
shalev396's picture
Add elevator-maintenance model: weights, model.py, Inference Endpoint handler, card
897d513 unverified
Raw History Blame Contribute Delete
2.68 kB
{
"model": "RandomForest",
"task": "binary-classification (failure state 10 min ahead from a 60-min window)",
"dataset": "Synthetic elevator IoT month (seeded generator, 44,640 minutely rows, 11 sensors)",
"split": "test",
"primary_metric": {
"name": "f1",
"value": 0.988235
},
"metrics": {
"f1": 0.988235,
"precision": 0.988235,
"recall": 0.988235,
"accuracy": 0.996983,
"pr_auc": 0.988996,
"roc_auc": 0.992683
},
"threshold": 0.37666666666666665,
"confusion_matrix": {
"tn": 1154,
"fp": 2,
"fn": 2,
"tp": 168
},
"selection": {
"rule": "highest validation F1 (ties: validation PR-AUC); threshold = max F1 on validation",
"val": {
"f1": 0.988764,
"precision": 0.988764,
"recall": 0.988764,
"accuracy": 0.996983,
"pr_auc": 0.981391,
"roc_auc": 0.996286
}
},
"comparison": {
"RandomForest": {
"val_f1": 0.988764,
"val_pr_auc": 0.981391,
"threshold": 0.376667,
"f1": 0.988235,
"precision": 0.988235,
"recall": 0.988235,
"accuracy": 0.996983,
"pr_auc": 0.988996,
"roc_auc": 0.992683,
"params": 10408,
"train_time_s": 2.9
},
"CNN-1D": {
"val_f1": 0.911681,
"val_pr_auc": 0.821253,
"threshold": 0.188755,
"f1": 0.958084,
"precision": 0.97561,
"recall": 0.941176,
"accuracy": 0.989442,
"pr_auc": 0.938044,
"roc_auc": 0.995726,
"params": 52993,
"train_time_s": 64.5
},
"LSTM": {
"val_f1": 0.9375,
"val_pr_auc": 0.954144,
"threshold": 0.166041,
"f1": 0.932927,
"precision": 0.968354,
"recall": 0.9,
"accuracy": 0.983409,
"pr_auc": 0.920017,
"roc_auc": 0.990769,
"params": 23681,
"train_time_s": 146.1
}
},
"examples": {
"failing": 0.813333,
"degrading": 0.92,
"healthy": 0.003333
},
"data": {
"n_train": 6236,
"n_val": 1326,
"n_test": 1326,
"n_failure_train": 413,
"n_failure_val": 178,
"n_failure_test": 170,
"n_rows": 44640,
"n_features": 63,
"n_classes": 2,
"window": 60,
"label_lag": 10,
"stride": 5,
"synthetic": true
},
"params": 10408,
"params_note": "RandomForest size = total tree nodes (Keras rows: weights)",
"train_time_s": 248.0,
"device": "cpu",
"smoke": false,
"source": "retrained 2026-09-25 with training/ (local CPU) on the synthetic month",
"versions": {
"sklearn": "1.9.1",
"numpy": "2.5.3",
"pandas": "3.0.6",
"joblib": "1.6.0",
"tensorflow": "2.21.0",
"keras": "3.15.1"
},
"trained_at": "2026-09-25"
}