Table 3 Comparison of HAMF Against Existing MLOps Frameworks.
From: Hybrid MLOps framework for automated lifecycle management of adaptive phishing detection models
Metric | HAMF | PhishBench | SageMaker | Kubeflow+MLflow |
|---|---|---|---|---|
Drift Detection Latency (s) | 18 | — | \(92 \pm 4.3\) | Manual |
Post-Drift Accuracy (%) | 99.52 ± 0.11 | 93.40 ± 0.29 | 96.10 ± 0.18 | 95.80 ± 0.23 |
p-value (vs HAMF) | — | <0.001 | 0.004 | 0.006 |
Fairness Violation (\(\Delta DP\)) | 0.03 ± 0.01 | 0.19 ± 0.03 | 0.08 ± 0.01 | 0.11 ± 0.02 |
SHAP Interpretability Retained | \(\checkmark\) | — | ✗ | \(\blacktriangle\) |
Automatic Feature Substitution | \(\checkmark\) | ✗ | ✗ | ✗ |
Fairness-Aware Retraining Pipeline | \(\checkmark\) | ✗ | ✗ | \(\checkmark\) |
