Future Failure Model Report
This report demonstrates the temporal training workflow for future-failure prediction.
Important: if the input is synthetic, these metrics are not real operational performance claims.
Run Configuration
- Input:
reports/synthetic_future_log.csv - Horizon days:
7 - Test start:
2026-03-01T00:00:00Z - Artifact:
artifacts/future_failure_model.joblib
Summary
| metric | value |
|---|---|
| train_rows | 708.0 |
| test_rows | 372.0 |
| roc_auc | 0.5812 |
| selected_threshold | 0.35 |
| accuracy | 0.7849 |
Confusion Matrix
| index | pred_no_failure | pred_failure |
|---|---|---|
| true_no_failure | 285 | 50 |
| true_failure | 30 | 7 |
Features
machine_typerun_hourssensor_1sensor_2sensor_3sensor_4sensor_5maintenance_type
Warnings
- Use real timestamped asset data before making maintenance claims.
- Synthetic data demonstrates mechanics only.
- Feature values must be available at or before prediction_timestamp.