Table 1

Dimension	Non-Hybrid Approaches (Heuristic / Optimization / ML)	Hybrid Models (Metaheuristic + ML / Multi-Hybrid)	Observations
Accuracy in Code Smell Detection	ML approaches (CodeT5, RefT5) improved detection accuracy by 25–30% by Armijo et al. [22].	Maini et al. [20] (Optimized Refactoring Sequence using PSO + DL): reported +30% accuracy improvement compared to rule-based.	Hybrid achieves higher precision, especially for complex systems.
Maintainability Improvement	Heuristics improved maintainability index by 20–25% by Kannangara et al. [23]. NSGA-II optimization improved modularity by 35% by Mkaouer et al. [24].	Maini et al. [19] HSHEP achieved +42% maintainability gain across benchmarks; Maini et al. [20] further showed enhanced scalability in OO systems.	Hybrid’s superior maintainability outcomes.
Effort Reduction	Decision Tree Forest sequencing reduced effort by 94.9% by Tarwani et al. [25].	Maini et al. [20] hybrid sequencing approach reduced sequencing search cost by ~15% vs standalone optimization.	Hybrids are less radical than some ML-only models but provide balanced efficiency + quality gains.
Scalability	Heuristics limited to small projects; NSGA-II scales better but with high computation.	Maini et al. [19] HSHEP successfully scaled sequencing to industrial-scale systems (JHotDraw, JFreeChart, Xerces, JEdit, Gantt Project) with stable results.	Hybridization (as shown in HSHEP) enables scalability while preserving quality.
Statistical Validation	Regression/classification (non-hybrid) reached R² = 0.877, F1 = 0.882 by Badru et al. [26].	Maini et al. [20] PSO+DL: MAPE reduced from 24.8% → 9.7%, p < 0.001 (statistically significant).	Hybrid models, especially in Ritika Maini’s work, show rigorous empirical and statistical validation.
Limitations	Heuristics lack adaptability; optimization needs parameter tuning; ML lacks explainability.	HSHEP & PSO+DL face complexity in setup, higher runtime cost, and reduced transparency.	Non-hybrids easier to implement; hybrids more powerful but harder to operationalize.

Table 1: Comparative Analysis: Non-Hybrid vs. Hybrid Models

Tables at a glance

Table 1

Figures at a glance

Figure 1

Figure 2

FIGURE 1

Figure 1: Flowchart of methodology showing data collection → implementation → evaluation → comparative analysis

FIGURE 2

Figure 2: Bar chart comparing heuristic, optimization, and ML approaches across maintainability, complexity reduction, and automation effort.