A hybrid Classifier using Weighted ELM with Biogeography Based Optimization for Hepatitis Diagnosis

Hepatitis is a liver disease caused by viral infections. An early diagnosis of Hepatitis is necessary to avoid mortality rate. In this research, the hybrid approach Weighted Extreme Learning Machine with Biogeography Based Optimization (WELM-BBO) is designed to diagnose Hepatitis. The machine learning model Weighted ELM has good generalization and efficient learning capacity. For imbalanced dataset, the classification algorithms always produce less accuracy. Since, WELM assigns distinct weights to minority and majority classes. The random initialization of the input parameters in WELM such as input weight and bias produce inconsistent and inaccurate classification results and degrade the classification performance. Hence, the Weighted ELM algorithm is hybridized with Biogeography Based Optimization (BBO) to find efficiently the optimum parameters such as input weight and bias. The proposed system is measured over hepatitis dataset. The classification efficiency of the proposed model is analyzed using the statistical parameters sensitivity, specificity, accuracy, and G-mean. The computational result clearly proved that WELM-BBO method yielded an accuracy of 86.96%, which is superior to other methods such as WELM, ELM, IBO-WELM, ACOR-WELM, BAT-ELM.