Heart Disease Risk Prediction using K-Nearest Neighbor: A Study of Euclidean and Cosine Distance Metrics

Cardiovascular disease, with an annual fatality rate of 17.9 million individuals, poses a significant global health crisis. Swift and precise diagnosis is paramount, particularly in regions with limited access to advanced medical technology. This paper employs machine learning techniques to predict heart disease, emphasizing hyperparameter optimization for heightened accuracy. Utilizing a diverse dataset to address issues like duplicate data, we conduct a comparative analysis of Euclidean and Cosine distances within the K-Nearest Neighbor (KNN) algorithm, filling a research gap and providing a comprehensive evaluation of these distance metrics. Label Encoding is applied for categorical variables, and KNN is implemented with varying neighbor counts $(1,3,5,7,9)$ using both Euclidean and Cosine distances. Evaluation metrics, including accuracy, precision, recall, and F1 score, gauge model performance. Our experiments reveal the highest accuracy of 90.761% when KNN is configured with 7 neighbors and the Cosine distance metric, surpassing the Euclidean distance. This underscores the criticality of careful distance metric selection in heart disease prediction models. The study offers insights into predictive modeling in healthcare, emphasizing the significance of distance metric choice. To streamline the abstract, omitting details related to future research and maintaining brevity is recommended.

I. Introduction

The escalation of cardiovascular disease to a formidable global health crisis is undeniable. Recent years have witnessed its emergence as a leading cause of mortality. According to the World Health Organization (WHO), this dire situation manifests in a staggering annual fatality rate of 17.9 million individuals succumbing to cardiovascular diseases. These alarming statistics underscore the pervasive and far-reaching impact of cardiovascular diseases on the landscape of global health [1]. Heart disease, a prevalent ailment primarily afflicting individuals in their middle and later years, frequently leads to fatal complications, accounting for a staggering one-third of all global fatalities [2]. In the contemporary healthcare environment, the diagnosis and treatment of heart disease remain notably formidable, especially in regions with limited access to advanced medical technology and a shortage of healthcare professionals. However, it is crucial to acknowledge that the accuracy of diagnosis and the timely implementation of medical interventions have the capacity to significantly mitigate the impact of heart disease and, in turn, preserve numerous lives. [3].