Framework Evaluation for Energy-Efficient Resource Allocation in Industrial IoT Using Hybrid Swarm Optimization: Comparative Analysis by Integrating Sand Cat Swarm Optimization and Moth Flame Optimization with SVM-Based Path Quality Assessment
Industrial Internet of Things (IIoT) environments generate enormous volumes of data that must be routed through cloud infrastructure with strict constraints on energy consumption, network latency, and service reliability. Existing resource allocation frameworks often rely on single-objective heuristics that fail to balance competing performance objectives across large-scale virtual machine (VM) deployments. The proliferation of heterogeneous IoT sensor nodes, each with distinct energy profiles and trust characteristics, further complicates optimal path selection in multi-hop cloud networks. These challenges demand intelligent optimization frameworks capable of simultaneously minimizing routing cost, energy usage, and latency while maximizing throughput and node trustworthiness.This paper proposes a novel hybrid swarm intelligence framework for energy-efficient resource allocation in Industrial IoT networks. The framework integrates two biologically inspired metaheuristic algorithms: Sand Cat Swarm Optimization (SCSO), which mimics the vibration-based hunting behaviour of sand cats to perform global exploration across candidate routing paths, and Moth Flame Optimization (MFO), which employs logarithmic spiral movement toward optimal flame positions to execute precise local refinement. The two-phase pipeline feeds the entire SCSO candidate population directly into the MFO refinement stage, enabling the hybrid to escape local optima that afflict standalone algorithms. A Support Vector Machine (SVM) classifier acts as a quality gate, filtering paths whose average trust score falls below 0.5 or whose average network latency exceeds 60 milliseconds. The framework is evaluated on a real-world dataset comprising 500 virtual machine nodes recorded at five-minute intervals, with 16 normalized feature dimensions including CPU utilization, memory usage, network throughput, latency, power consumption, response time, and SLA violation rate. Experimental results demonstrate that the Hybrid SCSO-MFO achieves a path efficiency of 95.3% at 300 nodes, representing improvements of 6.4 percentage points over GNN-based intrusion detection methods and 10.6 percentage points over dynamic graph transformer approaches. Energy consumption along optimal paths is reduced by 18.7% compared to standalone MFO and by 26.4% compared to standalone SCSO. Convergence is achieved 34% faster than MFO alone. These results confirm that the proposed framework delivers superior energy efficiency, network reliability, and scalability for IIoT cloud deployments.