Prediction of Factors for Patients with Hypertension and Dyslipidemia Using Multilayer Feedforward Neural Networks and Ordered Logistic Regression Analysis: A Robust Hybrid Methodology

Abstract

Background: Hypertension is characterized by abnormally high arterial blood pressure and is a public health problem with a high prevalence of 20%–30% worldwide. This research combined multiple logistic regression (MLR) and multilayer feedforward neural networks to construct and validate a model for evaluating the factors linked with hypertension in patients with dyslipidemia. Methods: A total of 1000 data entries from Hospital Universiti Sains Malaysia and advanced computational statistical modeling methodologies were used to evaluate seven traits associated with hypertension. R-Studio software was utilized. Each sample's statistics were calculated using a hybrid model that included bootstrapping. Results: Variable validation was performed by using the well-established bootstrap-integrated MLR technique. All variables affected the hazard ratio as follows: total cholesterol (β1: −0.00664; p < 0.25), diabetes status (β2: 0.62332; p < 0.25), diastolic reading (β3: 0.08160; p < 0.25), height measurement (β4: −0.05411; p < 0.25), coronary heart disease incidence (β5: 1.42544; p < 0.25), triglyceride reading (β6: 0.00616; p < 0.25), and waist reading (β7: −0.00158; p < 0.25). Conclusions: A hybrid approach was developed and extensively tested. The hybrid technique is superior to other standalone techniques and allows an improved understanding of the influence of variables on outcomes.