An Unsupervised Machine Learning Approach for Heart Disease Prediction

Cardiovascular diseases (CVDs) persist as a primary cause of mortality on a global scale, necessitating effective prediction methods. This study presents a novel modelling approach utilizing the Self-Organizing Map (SOM), an unsupervised machine learning approach, for CVD prediction. The SOM model employed to analyze and cluster patient data based on intrinsic patterns without predefined labels, enabling the identification of high-risk individuals. Results demonstrated that the SOM model effectively differentiates between healthy and at-risk patients, offering a robust approach for early detection. The SOM presented to be the best predictive model for the CVD data prediction with the highest accuracy (84.44%); precision (89.13%); recall (82.00%); F1-score (85.42%). The U-Matrix of SOM as a visualization tool provided insightful representations that highlighted two discernible clusters, effectively illustrating the health status of individuals with similar health conditions concerning CVD. This facilitates a better understanding and management of CVD. The research findings suggested the potential of integrating SOM into clinical workflows, offering a potent tool for healthcare professionals in the fight against CVD.