Ensemble Mutation Weight Convolutional Neural Network (Emwcnn) Classifier For Gene Expression Microarray Data

Main Article Content

V. Kalaimani, Dr.R. Umagandhi,

Abstract

Gene expression data can reflect gene activities and physiological status in a biological system at the transcriptome level. Gene expression data typically includes small samples but with high dimensions and noise. Hybrid Ensemble Feature Selection (HEFS) system is introduced recently for solving feature selection issue. In the recent work, K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Recursive Neural Networks (RNN) classifiers have been introduced for classification. These Machine Learning (ML) techniques are results in low accuracy and performance metrics because of single classifier. Thus Deep Learning (DL) architecture, achieved better results than using ML algorithms in terms of testing accuracy and performance metrics. In this paper, Ensemble Mutation Weight Convolutional Neural Network (EMWCNN) classifier is introduced for classification of gene expression data. To address the performance degradation of single ML classifiers, the construction of an EMWCNN classifier for multi‐platform fusion is proposed for gene expression data. In the training stage, multiple CNNs are trained as weak learners and fine‐tuned to minimize the weighted error. Meanwhile, optimum weights are selected via adaptive mutation operator. Subsequently, combine the predictions of multiple CNNs to achieve a boosting‐based prediction in the reference stage. EMWCNN method proves to be very efficient in classification of gene expression from microarray data, as it involves the process of considering opinion from multiple base classifiers, as opposed to the single classifier method.  Experimentation is carryout on four Gene Expression Microarray (GEM) datasets (Prostate cancer, Small Round Blue Cell Tumors (SRBCT), Leukemia, and Lymphoma). Experimental results verify that the proposed EMWCNN method shows improved results with respect to precision, recall, accuracy and Area UnderCurve (AUC) when compared to conventional classifiers.

Article Details

Section
Articles