Feature selection and risk prediction for patients with coronary artery disease using data mining

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Coronary artery disease (CAD) is a leading global cause of mortality.
• Early risk prediction of CAD enables timely interventions and reduces mortality rates.
• Existing studies often use extensive datasets, potentially leading to reduced prediction model performance due to data redundancy.

Purpose of the Study:

• To identify significant features for building accurate CAD risk prediction models.
• To enhance the performance of machine learning models for CAD risk stratification.
• To address data imbalance issues in cardiovascular datasets.

Main Methods:

• Feature selection using Chi-squared test, recursive feature elimination, and Embedded Decision Tree.
• Application of Synthetic Minority Over-sampling Technique (SMOTE) to handle imbalanced datasets.
• Development and evaluation of eight machine learning algorithms on Acute Coronary Syndrome (ACS) datasets from NCVD Malaysia.

Main Results:

• Identification of significant features crucial for CAD risk prediction.
• Implementation of SMOTE effectively addressed dataset imbalance.
• Top-performing prediction models achieved an Area Under the Curve (AUC) exceeding 90%.

Conclusions:

• Feature selection significantly improves the performance of CAD risk prediction models.
• Machine learning models, when optimized with feature selection and SMOTE, demonstrate high accuracy in predicting CAD risk.
• The findings support the development of more effective, data-driven tools for early CAD detection and management.