Chaotic Adaptive Particle Swarm Optimization and Quantum-Inspired Genetic Algorithm for Robust Feature Selection in IoT Intrusion Detection

The exponential growth of IoT data necessitates efficient feature selection for managing high-dimensional datasets while ensuring optimal classification performance. Traditional algorithms face premature convergence, local optima stagnation, and computational inefficiency. To address these, we used the following three advanced techniques. Chaotic Adaptive Particle Swarm Optimization (CAPSO) which integrates chaotic maps to improve exploration and diversity, avoiding premature convergence. It achieves 40%-60% feature reduction, 97%-98% accuracy, and 2%-3% higher accuracy scores than standard PSO. Hybrid Quantum Genetic Algorithm (HQGA) employs quantum principles to escape local optima, achieving 50%-65% feature reduction, 2% higher accuracy than CAPSO, and 10%-15% lower computational cost. Deep Learning-Assisted Chaotic Sparrow Search Algorithm (DLCSSA) combines chaotic maps with SSA and deep learning (e.g., transformer-based models) to guide the fitness function. It achieves 45%-55% feature reduction, 98%-99% accuracy, and 20% computational savings compared to standard SSA. By integrating chaos theory, quantum mechanics, and deep learning, these state-of-the-art methods significantly enhance performance, reduce computational costs, and improve robustness in IoT data analytics.