Investigation on breast cancer classification of snp genomics data using optimization based feature selection and deep learning algorithms

Abstract

Breast Cancer (BC) incidences are higher in women, and has been on the rise recently across the world. BC is represented by depiction of divergent genes inducing tumors with heterogeneous morphology and belligerence, generating different clinical symptoms. Individual differences in BC disease susceptibility and severity are further a result of these genetic polymorphisms. Single Nucleotide Polymorphism (SNP) is critical human ailments that have been found utilising Machine Learning (ML) techniques. SNP detection and the classification of healthy patients both provide considerable problems. It becomes quite difficult to identify and classify features from a dataset. Feature selections (FS) algorithms discover and eliminate unnecessary or redundant features resulting in reducing dimensionalities of datasets. Nondeterministic Polynomial (NP) occurs in traditional algorithms. Swarm Intelligence (SI) algorithms can handle NP hard problems. Bio-Inspired Hybrid Ensemble Feature selections (BIHEFS) algorithm are used on individual feature subsets for better approximations i.e. selections of optimal feature subsets for BC diagnostics. DL (DL) techniques have been introduced recently for classification of healthy and sick individuals based on SNP genomic data. Ensemble DL (EDL) introduced this work combines results of multiple individual models to enhance classifier performances. Three major contributions have been made to this work for BC diagnosis. newline