Cancer classification in high dimensional microarray gene expressions by feature selection using eagle prey optimization.

Microarray gene expression data have emerged as powerful tools in cancer classification and diagnosis. However, the high dimensionality of these datasets presents significant challenges for feature selection, leading to the development of various computational methods. In this paper, we utilized the Eagle Prey Optimization (EPO), a novel genetically inspired approach for microarray gene selection in cancer classification. EPO draws inspiration from the remarkable hunting strategies of eagles, which exhibit unparalleled precision and efficiency in capturing prey. Similarly, our algorithm aims to identify a small subset of informative genes that can discriminate between cancer subtypes with high accuracy and minimal redundancy. To achieve this, EPO employs a combination of genetic mutation operator with EPO fitness function, to evolve a population of potential gene subsets over multiple generations. The key innovation of EPO lies in its incorporation of a fitness function specifically designed for cancer classification tasks. This function considers not only the discriminative power of selected genes but also their diversity and redundancy, ensuring the creation of compact and informative gene subsets. Moreover, EPO incorporates a mechanism for adaptive mutation rates, allowing the algorithm to explore the search space efficiently. To validate the effectiveness of EPO, extensive experiments were conducted on several publicly available microarray datasets representing different cancer types. Comparative analysis with state-of-the-art gene selection algorithms demonstrates that EPO consistently outperforms these methods in terms of classification accuracy, dimensionality reduction, and robustness to noise.