Application of Horse Herd Optimization Algorithm for medical problems
Niloufar Mehrabi, Elnaz Pashaei · 2021 International Conference on INnovations in Intelligent SysTems and Applications (INISTA) · 2021
A vast impact concerning cancer research has been made by researchers analyzing the expression of thousands of genes at the same time and relating them to clinical phenotypes by DNA microarray technology. The large dimensions of gene expression microarray datasets, combined with a lack of samples and irrelevant or noisy genes, make classification tasks more difficult The purpose of gene selection is to identify the most informative genes in order to enhance prediction results. The Horse Herd Optimization Algorithm (HOA) is a novel swarm intelligence algorithm for solving continuous optimization problems, which simulate the behavior of a herd of horses. This study introduces a binary version of the Horse Herd Optimization Algorithm (BHOA) for solving discrete problems in biological data. Moreover, this paper introduces a novel hybrid gene selection framework that combines BHOA and a minimum Redundancy Maximum Relevance (mRMR) method. mRMR is applied as a filter approach to reduce noisy and irrelevant genes as well as to reduce the high computational cost that swarm intelligence algorithms generally suffer. The proposed approach’s performance was examined using four well-known datasets: DLBCL, Colon, SRBCT, and Leukemia. According to the experimental results, the hybrid approach performed better than all the other methods regarding both accuracy and the number of selected genes.