Table 9 Summary of Key Findings Aligned with Research Questions.
From: Hybrid MLOps framework for automated lifecycle management of adaptive phishing detection models
Research question | Empirical evidence | Interpretive significance |
|---|---|---|
RQ1: Performance under drift | Drift recovery in 18s; feature substitution restored F1 \(\approx\) 0.995 | Resilience to adversarial changes |
RQ2: Subsystem contributions | Ablation confirmed SHAP/drift modules critical to stability (F1-drop: 0.0135; p<0.01) (§ 6.4) | Explainability and fairness are integral |
RQ3: Benchmark comparison | Outperformed baselines in fairness (\(\Delta _{\textrm{DP}}<\) 0.08) and drift latency (18s vs. 92s and a 300s simulated manual response). | Establishes novelty beyond existing frameworks. |
RQ4: Scalability and fairness | Maintained \(\Delta _{\textrm{DP}} \le 0.03\) at 2.3k RPS; 41.6ms p99 latency ( §§ 6.6–6.7) | Ethics and speed achievable at scale |
