Clinical Dataset Classification Using Feature Ranking And Satin Bower Bird Optimized SVMs

Abstract

Abstract A clinical decision support system is a computer-based system that is designed to assist healthcare providers with clinical decision-making by analyzing electronic health records and other healthcare information systems to provide real-time support to clinicians at the point of care. A novel classification framework for clinical datasets in which the relevant features are selected by ranking them based on Fisher’s Score and a wrapper-based Satin Bower Bird Optimization algorithm with the combination of accuracy, G-mean and F-Score measured by support vector machine (SVM) as the fitness function is proposed. The classification is performed using an SVM classifier in which the hyperparameters of the SVM classifier are optimized using the Satin Bower Bird Optimization that improves the classification performance. In the context of statistical analysis, the research undergoes a non-parametric Friedman Test. This study selects relevant attributes from three clinical datasets from the Machine Learning Repository maintained by the University of California Irvine and achieved an accuracy of 86% for the Breast Cancer Wisconsin (Diagnostic) dataset, 89% for the Diabetic Retinopathy Debrecen dataset, and 91% for the EEG Eye State dataset respectively. When compared with other machine learning classifiers the proposed approach performed well with feature selection compared with other machine learning classifiers.