Computational intelligence approach for classification and risk quantification of metabolic syndrome / Habeebah Adamu Kakudi

• Main Author: Habeebah Adamu , Kakudi
• Format: Thesis
• Published: 2019
• Subjects: QA75 Electronic computers. Computer science; QA76 Computer software
• Online Access: http://studentsrepo.um.edu.my/14653/1/Habeebah.pdf; http://studentsrepo.um.edu.my/14653/2/Habeebah.pdf; http://studentsrepo.um.edu.my/14653/

Metabolic Syndrome (MetS) is clinically defined as the presence of three out of the following five abnormalities - ihyperglycaemia, raised waist circumference, low High- Density Lipoprotein-Cholesterol, ihypertriglyceridaemia and hypertension. MetS places individuals at an unhealthy disadvantage and iis associated with an increased risk of non-communicable diseases such as cardiovascular disease and diabetes. Currently used non-clinical methods are not able to diagnose the risk of MetS in patients that fall very close to the clinically defined threshold. Therefore, the aim of this study is to propose and develop a novel non-clinical technique for the early risk quantification and classification of MetS refered to as genetically optimized Bayesian adaptive resonance theory mapping (GOBAM). Genetic Algorithm(GA) is used to optimize the order of sequence of the input sample and the parameters of the Bayesian ARTMAP (BAM). The "Cohort study on clustering of lifestyle risk factors and understanding its association with stress on health and well-being among school teachers in Malaysia" (CLUSTer) dataset was used to compare the performance of the proposed Genetically Optimised Bayesian ARTMAP (GOBAM) model and three other classic Adaptive Resonance Theory Mapping (ARTMAP) models –Genetic Algorithm Fuzzy ARTMAP (GAFAM), Fuzzy ARTMAP (FAM), and Bayesian ARTMAP (BAM). GOBAM achieved higher of area under the receiver operating curve, sensitivity, specificity, positive predictive value, negative predictive value, and Fscore performance metrics of 91.45%, 96.3%, 88.3% , 98.32% , 85.71% , and 96.41% respectively. The proposed GOBAM model was able to diagnose the risk of MetS efficiently with borderline MRF measurements, by utilising a novel risk prediction index that ranged between 0 and 1.