GSVMA: A Genetic Support Vector Machine ANOVA Method for CAD Diagnosis

Area of Science:

• Cardiology
• Machine Learning
• Data Science

Background:

• Coronary artery disease (CAD) is a leading cause of cardiovascular mortality globally.
• Angiography is the standard diagnostic tool for CAD but is expensive and has side effects.
• Machine learning presents a promising alternative for accurate CAD diagnosis.

Purpose of the Study:

• To introduce a novel hybrid machine learning model, the genetic support vector machine and analysis of variance (GSVMA), for CAD diagnosis.
• To evaluate the performance of the GSVMA model against other machine learning methods using a benchmark dataset.

Main Methods:

• Developed a hybrid GSVMA model integrating genetic optimization algorithms with Support Vector Machine (SVM) using analysis of variance (ANOVA) as a kernel function.
• Employed a genetic algorithm for feature selection on the Z-Alizadeh Sani dataset.
• Compared GSVMA performance with linear SVM (LSVM) and LIBSVM with radial basis function (RBF).

Main Results:

• The GSVMA hybrid method achieved the highest accuracy of 89.45% using 10-fold cross-validation.
• Feature selection identified 31 crucial features contributing to the model's performance.
• GSVMA demonstrated superior performance compared to LSVM and LIBSVM (RBF) on the Z-Alizadeh Sani dataset.

Conclusions:

• Combining SVM with genetic optimization algorithms significantly enhances diagnostic accuracy for CAD.
• The GSVMA method is a highly effective tool that outperforms existing methods for CAD diagnosis.
• This approach can facilitate more accurate and potentially cost-effective CAD diagnosis.