Bio-inspired weighed quantum particle swarm optimization and smooth support vector machine ensembles for identification of abnormalities in medical data

Abstract

Knowledge mining is an emerging field where various patterns, rules, etc. can be generated which helps us in the analysis of the result. Medical information systems in modern hospitals and medical institutions become larger and larger, it causes great difficulties in extracting useful information for decision support. Especially, when traditional manual data analysis has become inefficient and methods for efficient computer-based analysis are indispensable. Decision support system is a suitable tool for medical application which helps in the extraction of useful data. A new hybrid model, weighted-quantum particle swarm optimization (WQPSO) for data clustering in sequence with smooth support vector machine (SSVM) is proposed for classification. The parameters considered to evaluate the clustering methods are inter-cluster distance, intra-cluster distance, and validation index. Experiments are performed in MATLAB and the performance analysis, comparisons are made with real-world datasets that are retrieved from UC Irvine Machine Learning Repository. The proposed approach overcomes the drawbacks of existing algorithms in terms of accuracy, performance and complexity. To validate the proposed algorithm, experiments are conducted on two data sets, i.e., liver disorders dataset and Wisconsin Breast Cancer Diagnosis (WBCD) dataset. The accuracy of proposed WQPSO-SSVM classification methodology is 83.76% for liver disorder, 98.42% for WBCD, 95.21% of mammography mass data. Among the considered algorithms such as WPSO-SVM, fuzzy, K-means, and fuzzy C-means methods, WQPSO-SSVM is found to yield a better convergence and provide an improved optimal solution.