The Evaluation of Ordinal Regression Model's Performance Through the Implementation of Multilayer Feed-Forward Neural...

Area of Science:

• Cardiovascular Research
• Computational Statistics
• Machine Learning in Healthcare

Background:

• Hypertension (high blood pressure) is a prevalent global health issue, significantly increasing the risk of cardiovascular diseases (CVD), stroke, and kidney disorders.
• Co-occurrence of hypertension and dyslipidemia presents complex challenges in patient management and risk prediction.

Purpose of the Study:

• To develop and validate a novel hybrid model integrating bootstrapping, ordered logistic regression, and multilayer feed-forward neural networks (MLFFNN).
• To identify and analyze key factors associated with hypertension in patients who also have dyslipidemia.

Main Methods:

• Utilized advanced computational statistical modeling techniques with RStudio software.
• A hybrid model combining bootstrapping, ordered logistic regression, and MLFFNN was developed and validated.
• Data from 33 participants at Hospital Universiti Sains Malaysia were analyzed, divided into training and testing datasets.

Main Results:

• The hybrid model demonstrated good model fitting, evidenced by predicted mean square error (MSE), mean absolute deviance (MAD), and accuracy.
• The model exhibited superior predictive capability, outperforming traditional regression models in predicting MSE.
• Key predictors identified include marital status, smoking status, systolic blood pressure, fasting blood sugar, and high-density lipoprotein levels.

Conclusions:

• The developed hybrid model offers a robust and superior approach for analyzing hypertension and dyslipidemia comorbidities.
• The study highlights the effectiveness of integrating diverse computational statistical methods for enhanced predictive accuracy in clinical settings.
• Findings underscore the importance of identified risk factors in managing patients with both hypertension and dyslipidemia.