Tuning the Parameters of Weighted ELM using IWO and BAT Algorithm to Improve the Classification Performance

Abstract

Weighted Extreme Learning Machine (WELM) is one among the machine learning algorithms with extremely learning and good generalization capabilities. WELM handles the imbalanced dataset efficiently for assigning less weight to majority class and more weight to minority class. In general, the classification performance of WELM extremely depends on the parameters such as the input weight matrix, the value of bias and the number of hidden neurons and the weights associated with majority and minority classes. The arbitrary selection of hidden biases and the input weight, WELM produces inconsistent result. In this paper, hybridization of WELM with Invasive Weed optimization and WELM with BAT algorithm are proposed to tune the parameters for WELM such as initial weight and hidden bias values. The proposed methodologies are called as WELM- IWO and WELM-BAT. The proposed methods are evaluated over three real world medical diagnosis problems such as Hepatitis, Diabetes and Thyroid diseases. The experimental results proved that one of the proposed methods WELM-IWO outperforms well on all three datasets.