A Quantum based Chaotic Evolutionary Strategy for Multimodal Feature Selection in Biomedical Data

Abstract

Medical datasets are inherently complex, often exhibiting high dimensionality, multimodal characteristics, and intricate interdependencies among features. These attributes pose significant challenges for traditional feature selection methods, which often struggle with overfitting, computational inefficiency, and difficulty in identifying multiple equally effective feature subsets (multimodal solutions). Most existing evolutionary algorithms tend to converge prematurely or fail to maintain diverse solutions, limiting their applicability to complex medical data. To address these limitations, this paper proposes a Quantum-Inspired Chaotic Evolutionary Feature Selection (QCEFS) approach, which integrates quantum-inspired operators, chaotic initialization, and a hybrid Particle Swarm Optimization (PSO)–Harris Hawk Optimization (HHO) framework. Chaotic initialization ensures wide and diverse exploration of the solution space from the outset, helping to avoid local optima. Quantum-inspired operators enable probabilistic representation and superposition of candidate solutions, promoting a dynamic and efficient search strategy. The hybrid PSO-HHO mechanism further enhances convergence reliability by balancing exploration and exploitation. Moreover, to effectively handle multimodal feature selection—where multiple feature subsets of same cardinality and similar classification accuracy exist—a speciation-based niching technique is employed. This encourages the algorithm to discover and preserve diverse, high-performing multiple feature subsets. Experiments on 21 real-world medical datasets demonstrate the superiority of the proposed method over eight state-of-the-art evolutionary feature selection algorithms. The results indicate notable improvements in classification accuracy, subset compactness, and convergence robustness. Statistical tests further confirm the significant performance of the proposed QCEFS algorithm.