Random local SVMs for classifying large datasets

Abstract

We propose a new parallel ensemble learning algorithm of random local support vector machines, called krSVM for the effectively non-linear classification of large datasets. The random local SVM in the krSVM learning strategy uses kmeans algorithm to partition the data into k clusters, followed which it constructs a non-linear SVM in each cluster to classify the data locally in the parallel way on multi-core computers. The krSVM algorithm is faster than the standard SVM in the non-linear classification of large datasets while maintaining the classification correctness. The numerical test results on 4 datasets from UCI repository and 3 benchmarks of handwritten letters recognition showed that our proposed algorithm is efficient compared to the standard SVM.