Comprehensive Review of Hybrid Feature Selection Methods for Microarray-Based Cancer Detection

Microarray data-based cancer detection advances early diagnosis and personalized medicine by utilizing gene expression data to develop comprehensive cancer profiles, measuring thousands of genes simultaneously. However, the inherent high-dimensional nature of microarray data introduces substantial challenges in data analysis and interpretation. To resolve these issues, gene selection techniques such as filter, wrapper, and embedded methods have been implemented to remove irrelevant genes and reduce the dimensionality of the data. Even with such usefulness, these methods are bound to restrictive elements individually that could compromise the precision of cancer detection systems. More recently, the focus of research has shifted to hybrid approaches that merge several feature selection techniques to mitigate the weaknesses of one method while maximizing the strengths of others. This paper offers an extensive review on feature selection techniques for microarray data and focuses on evaluating the performance of different hybrid methods as an important research gap. The research assesses various combinations of Filter, Wrapper, and Embedded techniques to determine how such hybrid approaches enhance classification accuracy. Hybrid approaches, those that integrate several techniques, have the ability to enhance diagnostic accuracy as well as improve understanding at the biological level. This paper provides a comparative evaluation of hybrid feature selection methods to enhance microarray-based cancer classification. It aims to guide researchers in choosing appropriate strategies that optimize the dataset analysis.