Feature Selection Technique Based on Neuro-Fuzzy-Rough Set for Cancer Classification Using Gene Expression Data
Nancy Gilbert, A. Balamurugan · SSRN Electronic Journal · 2016
The study of gene expression in cancer diagnosis is done by microarray technology. The description of microarray technology issues (to select the corrective set of genes and curse of dimensionality) and proposed technique to overcome these issues are the scope of this paper. Proposed technique helps to improve the performance of learning models by removing most irrelevant, redundant and noise features from the datasets through a hybrid feature selection method. In recent days many researches endeavor hybrid feature selection techniques which combined two or more feature selection techniques rather than using a single one. In this paper, we propose Neuro-fuzzy-rough set based hybrid feature selection (NFR) method, namely FPRS and ANFIS, which extracted top ranked genes and found more relevant genes as well as noisy genes from the microarray gene expression data to get better classification accuracy. This work is applied on four micro array datasets and the experimental results are revealed its efficiency and effectiveness of proposed technique. Finally, obtained salient genes are evaluated with two classification algorithms (kNN and SVM).