Predicting heart disease among males based on performance analysis using machine learning technology: an unsupervised feature selection approach

Abstract

Abstract—Machine learning algorithms play a vital role in the early intervention of heart disease by enabling accurate, data-driven insights. This study focuses on predicting heart disease among males using an unsupervised learning approach applied to the PTB-XL ECG dataset. After thorough preprocessing—including MinMaxScaler normalisation, Variance Thresholding, and PCA—K-Means Clustering was employed to detect hidden patterns in patient data. The optimal model with k=8 achieved a silhouette score of 0.82 and inertia of 5.09, outperforming Agglomerative and DBSCAN in both cohesion and interpretability. These results confirm the effectiveness of unsupervised learning in extracting clinically relevant patterns and highlight the importance of gender-specific modelling in heart disease prediction. The proposed framework offers a scalable and robust solution for early risk assessment and supports future development of intelligent, sex-aware diagnostic systems.