Abstract

Early risk factor detection is essential for managing and treating cardiovascular disease (CVD), a global health issue. Studies have shown that waist circumference (WC) and waist hip ratio (WHR) are better at identifying CVD than BMI. The study uses Random Forest (RF) machine learning to identify characteristics that affect WHR, an indication of CVD. Isfahan Cardiovascular Research Centre in Iran provided the dataset, which includes sex, family history, diabetes, WHR, smoking, systolic blood pressure, and total cholesterol. Statistical analyses employed Yates' correction and Pearson chi-squared tests. Modeling with RF yielded accuracy, balanced accuracy, sensitivity, specificity, PPV, NPV, and F1 score from performance metrics. Finally, variable significance values determined the dependent variable's most relevant variables. WHR and other variables are statistically significantly correlated. Random Forest machine learning predicts high WHR with high accuracy, sensitivity and specificity. The most important variables of the prediction model are female sex, smoking status and blood pressure ranges. In conclusion, the global burden of CVD and the necessity of early diagnosis are underlined. The role of WHR along with BMI and waist circumference in the assessment of cardiovascular risk is emphasised. The study concludes that the machine learning model can effectively predict high WHR, aid CVD risk management and facilitate personalised treatment plans. The results contribute to a better understanding of the factors influencing high WHR and can guide healthcare professionals in the comprehensive assessment and management of cardiovascular risks.

Key words: Cardiovascular disease, risk factor, waist hip ratio, machine learning