A STUDY ON A NEW APPROACH TO HYBRID REGRESSION MODELING: A CASE FOR DIABETES MELLITUS WITH DYSLIPIDAEMIA PATIENTS WHO VISITED HOSPITAL USM

Abstract

Background and Objective: The incidence of type 2 diabetes has been steadily rising over the past few decades, which has contributed significantly to the rise in the prevalence of diabetes (DM). Statistics from the World Health Organization show that more than 422 million adults worldwide had diabetes in 2014, and an ongoing rise in DM prevalence is anticipated. This study aims to create a method that can use to predict and manage diabetes cases in light of the importance of statistical modeling in diabetes. Decision trees and ordinal regression were the two methods used in this study. With some modification and extension, both methods will be harmonized in the R syntax. Materials and Methods: In this paper, we developed a method for analyzing decision trees using R syntax and embedding classification predictions. The classification for prediction with accuracy will indicate a successful classification analysis. This study illustrated the development method using diabetes data consisting of one thousand observations. Before further testing, the clinical relevance and significance of each preselected variable will be assessed. The decision tree will be used to evaluate nine variables. The selected variables are body mass index, total cholesterol, diabetes status, glucose reading, high-density lipoprotein, patient height, hip circumference, hypertension status, smoking status, and triglycerides. The classification obtained will be used as an input for the ordinal regression modeling. Result: It has been discovered that the status of diabetes can be determined by the level of glucose during fasting, which is consistent with the most recent research that has been published. one variable was chosen and used for the input of the ordinal regression. The suggested variables will apply to the ordered logistic regression, and the developed syntax will be used to assess the goodness of measurement and the significance level is set at a 0.05 level. Conclusion: Our proposed method achieves the highest level of forecasting precision possible. The methodology offers a precise evaluation of the fit of the final model. The superior performance of the model resulted in improved outcomes and efficient decision-making management.