Metaheuristics for Machine Learning

4Enhancing Feature Selection Through Metaheuristic Hybrid Cuckoo Search and Harris Hawks Optimization for Cancer Classification

Abrar Yaqoob¹*, Navneet Kumar Verma¹, Rabia Musheer Aziz² and Akash Saxena³

¹School of Advanced Science and Language, VIT Bhopal University, Kothrikalan, Sehore, India

²State Planning Institute (New Division), Planning Department Lucknow, Utter Pradesh, India

³School of Engineering and Technology, Central University of Haryana, Mahendergarh, India

Abstract

Gene expression platforms offer vast amounts of data that can be utilized for investigating diverse biological processes. However, due to the existence of redundant and irrelevant genes, it remains challenging to identify crucial genes from high-dimensional biological data. To overcome this obstacle, researchers have introduced different feature selection (FS) methods. Developing more efficient and accurate FS techniques is essential to select important genes for the classification of complex biological information with multiple dimensions for many purposes. To tackle the difficulty of selecting genes in high-dimensional biological datasets, a novel approach called the Harris hawks optimization and cuckoo search algorithm (HHOCSA) is proposed for commonly used machine learning classifiers such as K-nearest neighbors (KNN), support vector machine (SVM), and naive Bayes (NB). The effectiveness of the hybrid gene selection algorithm was assessed using six commonly used datasets and compared to other ...