A Multiclass Classification of Cancer Data: Using a Kernel Based Clustering k-NN Support Vector Machine

Support vector machines (SVM) have been promising methods for classification because of their solid mathematical foundations which convey several salient properties that other methods hardly provide. However, despite of the prominent properties of SVM, they are not as favored for large-scale data as complexity of SVM is highly dependent on the size of a data set. Microarray gene expression data that usually have a large number of dimensions, over thousands of genes, and a small number of samples, e.g., a few tens of patients. This paper presents a noble and efficient approach, Dimensionality Reduction (T-test), followed by Clustering kNN Support Vector machines (CK-SVM), which is specially designed for handling very large data sets like Microarray gene expression data. The CK-SVM classifies by reflecting the degree of a training data point, as a support vector by using Gaussian function, with K-nearest neighbor (k-NN) and Euclidean distance measure. To add local control property a simple clustering scheme is implemented, before Gaussian functions are constructed for each cluster. In addition probabilistic SVM out puts are used for extending binary classification to multiclass classification, in a pair wise manner. In this paper a multiclass classification has been applied to cancer data, represented by SRBCT data set.