GAN-generated Synthetic Data and SVM-based Feature Selection for Improved Cardiovascular Disease Prediction

Early diagnosis requires cardiovascular disease forecasting. Past patient interests can improve machine learning predictions. This study uses a Generative Adversarial Network (GAN) to augment data to address the issue of insufficient datasets. Our objective is to use the GAN and feature selection (FS) to tackle the issue of class imbalance. The study assesses the efficacy of three different machine learning (ML) algorithms in predicting coronary artery disease (CAD). The UCI repository provided the Z-Alizadeh sani dataset for our analysis. A GAN was used to balance an imbalanced dataset. We used GAN to augment data and synthesise data to improve the models. Support Vector Machine (SVM) as feature selection tool was used to extract more effective attributes. We trained and tested CAD prediction using machine learning algorithms using 5-fold cross validation. We compared the machine-learning algorithm’s outcomes with and without GAN-based data enhancement and feature selection to evaluate its predicted accuracy. According to results GAN and feature selection improved accuracy, sensitivity, and specificity for all algorithms, suggesting that generated data and selected attributes can identify positive and negative cases.