Neuro-fuzzy ensembler for cognitive states classification

The functional magnetic resonance imaging (fMRI) is considered as a powerful technique for performing brain activation studies by measuring neural activities. However, the tons of voxels over time are posed a major challenge to neuroscientists and researchers for analyzing it effectively. The decoding of brain activities required fast, accurate, and reliable classifiers. In classification scenario, although machine learning classifiers have shown promising result, but the individual classifiers have their limitations. This paper proposes a method based on the ensemble of Neural Networks applying on fMRI data for cognitive state classification. The Neural Networks (NNs) classifier has been selected for ensembling. The Fuzzy Integral (FI) approach is used as an efficient tool for combining different classifiers. The classifiers ensemble technique performs better than the single base learner by reducing misclassification as well as both bias and variance. The proposed technique successfully classifies different cognitive states with high classification accuracy. The performance improvement while applying the ensemble technique as compared with the individual neural network strongly recommends the usefulness of the proposed approach.