Optimized Gene Selection Model for Accurate Classification of Microarray Gene Expression Data

In the realm of gene expression analysis, isolating significant genes from high-dimensional biological datasets remains a critical challenge, often hindered by redundancy and the presence of irrelevant features. To address this, we propose a novel hybrid gene selection algorithm that integrates Harris Hawk Optimization (HHO) with Cuckoo Search Algorithm (CSA), termed HHOCSA, synergized with the Support Vector Machine (SVM) classifier for the effective classification of biomedical data. This method is evaluated on the benchmark Lung Cancer dataset, demonstrating its superiority over traditional feature selection (FS) methods. The HHOCSA algorithm effectively identifies relevant gene subsets by leveraging a hybrid strategy, achieving notable improvements in classification accuracy, specificity, and sensitivity. Experimental results reveal that HHOCSA achieves mean classification accuracies of 100%, 96.70%, and 99.39% in Experiments 1, 2, and 3, respectively, outperforming mRMR, mRMR + HHO, and mRMR + CSA in all tested scenarios. The findings underscore the robustness and efficiency of HHOCSA in handling high-dimensional data, making it a valuable tool for bioinformatics and biotechnology researchers engaged in gene selection and classification tasks.