Voting based Extreme learning Machine with Spectral Coefficient Pruning for binary Classification

Abstract

Extreme Learning machine (ELM) is emerged as an efficient fast learning classifier for real valued classification problems. Voting Based ELM, V-ELM uses majority voting based ensembling technique to further improve the performance of ELM. V-ELM gives better performance at the cost of increased computational and memory requirement. This paper extends V-ELM by incorporating recently proposed spectral coefficient pruning technique, which reduces the aforementioned problems. The extended classifier is referred as Voting based ELM with Spectral coefficient Pruning, V-ELM_SP. Spectral coefficient pruning ensures that the component classifiers of pruned ensemble has both accurate and diverse classifiers. This work evaluates V-ELM_SP using various datasets available at Keel dataset repository. V-ELM_SP performs better than V-ELM for almost all evaluated datasets.