C-HMOSHSSA: Gene selection for cancer classification using multi-objective meta-heuristic and machine learning methods. (September 2019)

• Record Type: Journal Article
• Authors: Sharma, Aman; Rani, Rinkle
• Is Part Of: Computer methods and programs in biomedicine. Volume 178(2019)
• Page Start: 219
• Page End: 235
• Publication Date: 2019-09
• Subjects: Multi-objective optimization -- Spotted hyena optimizer -- Salp swarm algorithm -- Cancer classification -- Gene expression -- Machine learning -- Gene selection -- Feature selection; Medicine -- Computer programs -- Periodicals; Biology -- Computer programs -- Periodicals; Computers -- Periodicals; Medicine -- Periodicals; Médecine -- Logiciels -- Périodiques; Biologie -- Logiciels -- Périodiques; Biology -- Computer programs; Medicine -- Computer programs; Periodicals; Electronic journals; 610.28
• DOI: 10.1016/j.cmpb.2019.06.029
• Languages: English
• ISSNs: 0169-2607

Abstract

Highlights: A hybrid approach is proposed for gene selection using two powerful meta-heuristic approaches; SSA and MOSHO. The issues related due to high-dimensional data are addressed. Salp Swarm Algorithm (SSA) is used to maintain the diversity and MOSHO for maintaining faster convergence. There is no restriction of datasets for the proposed approach. It can be applied for different types of diseases. Four popularly used classifiers are trained on seven different high-dimensional datasets. Abstract: Background and objective: Over the last two decades, DNA microarray technology has emerged as a powerful tool for early cancer detection and prevention. It helps to provide a detailed overview of disease complex microenvironment. Moreover, online availability of thousands of gene expression assays made microarray data classification an active research area. A common goal is to find a minimum subset of genes and maximizing the classification accuracy. Methods: In pursuit of a similar objective, we have proposed framework (C-HMOSHSSA) for gene selection using multi-objective spotted hyena optimizer (MOSHO) and salp swarm algorithm (SSA). The real-life optimization problems with more than one objective usually face the challenge to maintain convergence and diversity. Salp Swarm Algorithm (SSA) maintains diversity but, suffers from the overhead of maintaining the necessary information. On the other hand, the calculation of MOSHO requires low computational efforts hence is used for maintaining the necessary information. Therefore, the proposed algorithm is a hybrid algorithm that utilizes the features of both SSA and MOSHO to facilitate its exploration and exploitation capability. Results: Four different classifiers are trained on seven high-dimensional datasets using a subset of features (genes), which are obtained after applying the proposed hybrid gene selection algorithm. The results show that the proposed technique significantly outperforms existing state-of-the-art techniques. Conclusion: It is also shown that the new sets of informative and biologically relevant genes are successfully identified by the proposed technique. The proposed approach can also be applied to other problem domains of interest which involve feature selection.