Modeling the Factors Associated with BMI among Type 2 Diabetes Mellitus Patients: A Hybrid Model Approach
Farah Muna Mohamad Ghazali
School of Dental Sciences, Health Campus, Universiti Sains Malaysia (USM), 16150 Kubang Kerian, Kota Bharu, Kelantan, Malaysia.
Wan Muhamad Amir W. Ahmad *
School of Dental Sciences, Health Campus, Universiti Sains Malaysia (USM), 16150 Kubang Kerian, Kota Bharu, Kelantan, Malaysia.
Mohamad Nasarudin Adnan
School of Dental Sciences, Health Campus, Universiti Sains Malaysia (USM), 16150 Kubang Kerian, Kota Bharu, Kelantan, Malaysia.
Norsamsu Arni Samsudin
School of Dental Sciences, Health Campus, Universiti Sains Malaysia (USM), 16150 Kubang Kerian, Kota Bharu, Kelantan, Malaysia.
Nor Azlida Aleng
Faculty of Ocean Engineering Technology and Informatics, Universiti Malaysia Terengganu (UMT), 21030 Kuala Nerus, Terengganu, Malaysia.
Nor Farid Mohd Noor
Faculty of Medicine, Universiti Sultan Zainal Abidin (Uni SZA), Medical Campus, Jalan Sultan Mahmud, 20400 Kuala Terengganu, Terengganu, Malaysia.
Mohamad Shafiq Mohd Ibrahim
Kulliyyah of Dentistry, International Islamic University Malaysia, IIUM Kuantan Campus, Jalan Sultan Ahmad Shah, Bandar Indera Mahkota, 25200 Kuantan, Pahang, Malaysia.
Nurul Hidayah Binti Shamsudin
Department of General Medicine, Hospital Tawau, Peti Surat 67, 91007 Tawau, Sabah, Malaysia.
Siddharthan Selvaraj
Faculty of Medicine, Universiti Sultan Zainal Abidin (Uni SZA), Medical Campus, Jalan Sultan Mahmud, 20400 Kuala Terengganu, Terengganu, Malaysia.
*Author to whom correspondence should be addressed.
Abstract
Background: Diabetes mellitus is a chronic illness that results in abnormally high blood sugar levels. It can result in a range of complications.
Objective: The purpose of this study is to present an ideal variable selection strategy utilizing proven Multiple Linear Regression (MLR) models and to validate the variable using Multilayer Perceptron Neural Network (MLP) models. This will validate a factor linked with body mass index (BMI) status in individuals with dyslipidemia and type 2 diabetes mellitus.
Materials and Methods: Thirty-nine patients were selected from Hospital Universiti Sains Malaysia (USM). Many variables, including BMI, gender, age, race, coronary heart disease status, waist circumference, alanine transferase, triglycerides, and dyslipidemia, were assessed in this retrospective analysis using advanced computational statistical modelling approaches. This study uses R-Studio software and syntax. Each sample's statistics were generated using a hybrid model combining bootstrap and multiple linear regression.
Results: R's statistical approach demonstrates that regression modelling is superior to R-squared performance. The hybrid model may better predict the outcome by separating the datasets into a training and testing set. The well-known bootstrap-integrated MLR technique was used to determine the validity of the variables. The eight variables examined in this case are gender ( : -2.329; p < 0.25), age ( : -0.151; p < 0.25), race ( : 2.504; p < 0.25), coronary heart disease status ( : -0.481; p < 0.25), waist circumference ( : 0.572; p < 0.25), alanine transferase ( : 0.002; p < 0.25), triglycerides ( : 0.046; p < 0.25), and dyslipidemia ( : 30.769; p < 0.25). There is a linear model that has a 9.019188 MSE.lm in this case.
Conclusion: This study will develop and extensively evaluate a novel hybrid approach combining bootstrapping and multiple linear regression. The R syntax for this procedure was chosen to ensure that the researcher comprehends the example completely. The statistical methods used to conduct this research study using R show that regression modelling is better than R-squared values for the predicted mean squared error. Thus, the study's conclusion shows that the hybrid model technique is superior. This vital conclusion helps us better understand the hybrid method's relative contribution to the result in this case.
Keywords: BMI, Modelling, multiple linear regression, predicted mean square error, type 2 diabetes mellitus