Disagreement Measure Based Ensemble of Extreme Learning Machine for Gene Expression Data Classification

Abstract

Selective ensemble learning has become a powerful tool for biological information analysis of gene expression data.In order to mining gene expression data better,we use ensemble of Extreme Learning Machine(ELM) to overcome the shortage that a single ELM is unstable in data classification.In this paper,we propose an algorithm,which is the dissimilarity ensemble based on disagreement measure of Extreme Learning Machines(D-D-ELM).First,we judge the dissimilarity of Extreme Learning Machines with disagreement measure.Then we remove the corresponding ELMs based on the average classification accuracy.At last,the rest ELMs are grouped into an ensemble classifier by the strategy of majority voting.This algorithm is applied on the data of gene expression Breast,Leukemia,Colon,Heart.The theoretical analysis and experiment are given and the statistical analysis on the experimental results demonstrates that D-D-ELM can achieve better classification accuracy with less number of ELMs.