Comparison of artificial neural network, logistic regression and discriminant analysis methods in prediction of metabolic syndrome.

摘要

Introduction: Artificial neural networks as a modern modeling method have receivedudconsiderable attention in recent years. The models are used in prediction and classification inudsituations where classic statistical models have restricted application when some, or all of theirudassumptions are met. This study is aimed to compare the ability of neural network models touddiscriminant analysis and logistic regression models in predicting the metabolic syndrome.udMaterials & Methods: A total of 347 participants from the cohort of the Tehran Lipid and GlucoseudStudy (TLGS) were studied. The subjects were free of metabolic syndrome at baseling according toudthe ATPIII criteria. Demographic characteristics, history of coronary artery disease, body massudindex, waist, LDL, HDL, total cholesterol, triglycerides, fasting and 2 hours blood sugar, smoking,udsystolic and diastolic blood pressure were measured at baseline. Incidence of metabolic syndrome afterudabout 3 years of follow up was considered a dependent variable. Logistic regression, discriminantudanalysis and neural network models were fitted to the data. The ability of the models in predictingudmetabolic syndrome was compared using ROC analysis and the Kappa statistic, for which, MATLABudsoftware was used. Results: The areas under receiver operating characteristic (ROC) curve for logisticudregression, discriminant analysis and artificial neural network models (15: 8: 1) and (15: 10: 10)udwere estimated as 0. 749, 0. 739, 0. 748 and 0. 890 respectively. Sensitivity of models wereudcalculated as 0. 483, 0. 677, 0. 453 and 0. 863 and their specificity as 0. 857, 0. 660, 0. 910 and 0. 844udrespectively. The Kappa statistics for these models were 0. 322, 0. 363, 0. 372 and 0. 712udrespectively. Conclusion: Results of this study indicate that artificial neural network modelsudperform better than classic statistical models in predicting the metabolic syndrome.