Classification of cognitive state using clustering based maximum margin feature selection framework

Over the past few years, the dimensionality of functional MRI (fMRI) effects the analysis of brain data. In the field of machine learning and statistical analysis, classification of objects plays a significant role. Machine learning classifiers are used to discover the class of new data points from a set of data points. The application of learning techniques on fMRI data alleviates to cognitive state classification. High dimensional fMRI data has thousands of features in the form of voxels. Choice of appropriate voxels as attributes is an ambitious activity for classification of cognitive state. A novel feature selection framework, a clustering based maximum margin through sparse constraints is proposed in this paper. The performance of the proposed method is examined on benchmark fMRI data set. K-means clustering and feature selection are merged into a consistent framework. L
21
norm regularization are executed to obtain relevant features from the data. The method provides few number of most appropriate features from StarPlus fMRI data. The obtained voxels are furnished as features to Naive Bayes classifier.