An Efficient Diagnosis System for Thyroid Disease Based on Enhanced Kernelized Extreme Learning Machine Approach

In this paper, we present a novel hybrid diagnosis system named LFDA-EKELM, which integrates local fisher discriminant analysis (LFDA) and kernelized extreme learning machine method for thyroid disease diagnosis. The proposed method comprises of three stages. Focusing on dimension reduction, the first stage employs LFDA as a feature extraction tool to construct more discriminative subspace for classification, the system switches from feature extraction to model construction. And then, the obtained feature subsets are fed into designed kernelized ELM (KELM) classifier to train an optimal predictor model whose parameters are adaptively specified by improving artificial bee colony (IABC) approach. Here, the proposed IABC method introduces an improved solution search equation to enhance the exploitation of searching for solutions, and provides a new framework to make the global converge rapidly. Finally, the enhanced-KELM (EKELM) model is applied to perform the thyroid disease diagnosis tasks using the most discriminative feature subset and the optimal parameters. The effectiveness of the proposed system is evaluated on the thyroid disease dataset in terms of classification accuracy. Experimental results demonstrate that LFDA-EKELM outperforms the baseline methods.