Heart Disease Diagnosis: a Comparative Study of Traditional and Quantum-Based Approaches Using Machine Learning

Abstract

Heart disease remains a primary global health issue, the leading cause of death worldwide. This highlights the need for diagnostic methods that are accurate, rapid, accessible, and costeffective. This study employed a clinical dataset from Kaggle, containing 500 observations with six variables: age, gender, blood pressure, cholesterol, heart rate, and a quantum pattern feature. Several supervised machine learning models were developed and compared to classify and predict heart disease. The data was preprocessed, balanced, and split into training and testing sets to ensure unbiased results. Models tested include decision trees, discriminant analysis, logistic regression, Naïve Bayes, support vector machines, linear classifiers, and k-nearest neighbors. Among these, the binary (GLM) logistic regression model achieved 94.72% cross-validation accuracy and 95.00% accuracy on the testing set. The linear discriminant model performed even better, reaching 97.50% accuracy on the testing set. Using feature selection, the number of variables was reduced, and with only one variable, the quantum pattern feature method achieved 97.50% accuracy. These results suggest that quantum-based diagnostics are highly effective. Based on this, the research team conducted a comparative study between standard traditional and quantum pattern feature methods for heart disease to evaluate each approach's accuracy and determine whether the conventional method can achieve comparable precision. The traditional method can achieve an accuracy of 77.50%. This study further aims to promote diagnostic strategies that are both feasible and implementable in real-world clinical settings.

Ähnliche Arbeiten