Clustering Techniques for Heart Attack Prediction: A Silhouette Score-Based Evaluation

Heart attacks, also known as myocardial infarctions, remain a major contributor to global mortality, necessitating early detection of individuals at risk.Traditional diagnostic methods often fail to uncover complex patterns in medical data.This study explores the use of unsupervised machine learning techniques (K-Means, K-Medoids, and Agglomerative Clustering) to segment patients based on cardiovascular risk factors and support heart attack prediction.A publicly available heart attack dataset was preprocessed using normalization and label encoding.The clustering outcomes were evaluated using the Silhouette Score.Among the three techniques, K-Medoids achieved the highest Silhouette Score, indicating more coherent and meaningful patient groupings.