DIFFERENTIAL EVOLUTION ALGORITHM AS FEATURE SELECTION FOR BIOMARKER DISCOVERY

Authors

  • S. A. M. Yusoff Faculty of Computer and Mathematical Sciences, UiTM Pulau Pinang
  • S. M. Zahari Faculty of Computer and Mathematical Sciences, UiTM Selangor
  • F. H. Mustafa Faculty of Computer and Mathematical Sciences, UiTM Pulau Pinang
  • N. A. Rahman Faculty of Computer and Mathematical Sciences, UiTM Kelantan
  • M. H. Abdullah Faculty of Electrical Engineering, UiTM Pulau Pinang

DOI:

https://doi.org/10.4314/jfas.v10i2s.77

Keywords:

differential evolution, feature selection, biomarker discovery, classification, bio-inspired.

Abstract

The advancement in mass spectrometry technique for proteomic studies has proliferated the discovery of biomarkers from quantitative proteomics pattern. High-throughput data for a given molecule can give rise to a series of inter-related and overlapping peaks in a mass spectrum. The spectrum suffers from high dimensionality data relative to small sample size. Feature selection techniques search parsimonious features through a learning model that exhibits the most accurate results. A computational technique that mimics survival and natural processing known as DE integrated with linear SVM classifier was proposed for feature selection. The comparisons have been made with PSO and ACO algorithms. The proposed feature selection of DE algorithm exhibited accuracy, sensitivity and specificity with 82.2, 80.0 and 84.0% on liver (HCC) datasets respectively and outperformed the PSO and ACO.

Downloads

Published

2018-05-31

Issue

Section

Research Articles