Predicting Heart Disease Risk Using Logistic Regression and Random Forest Models: Balancing Interpretability and Accuracy

Abstract

Cardiovascular disease, particularly heart disease, continues to be a major contributor to global mortality, making timely detection a critical priority. Traditional diagnostic approaches often identify the illness only after significant progression, underscoring the need for predictive tools that rely on clinical and demographic information. This study utilizes the Cleveland subset of the UCI Machine Learning Repository, which contains 270 patient records with 14 attributes, to explore predictive modeling for heart disease. We developed and assessed two methods: logistic regression and random forest. Logistic regression achieved an accuracy of 85% and an AUC of 0.918, offering transparency in interpreting risk factors such as ST depression, abnormal vessel counts, and thallium test outcomes. The random forest model delivered a comparable accuracy of 85% with an AUC of 0.906, highlighting similar predictors while capturing nonlinear patterns within the data. Our findings indicate that combining interpretable models with machine learning techniques provides a balanced and reliable strategy to support early heart disease detection and improve clinical decision-making.