|
Abstract
|
Cardiovascular disease (CVD) ranks among the top causes of mortality globally, underscoring the urgent necessity for advanced predictive models to enhance early detection and preventative measures. In this direction, this study investigates the performance of five well-established deep learning (DL) models, namely Multilayer Perceptron (MLP), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and Autoencoder in predicting CVD using a diverse patient dataset. To tackle the prevalent class imbalance issue in medical datasets, we introduce an enhanced Synthetic Minority Over-sampling Technique (SMOTE). This innovative technique enhances traditional SMOTE by incorporating feature correlations to produce more realistic synthetic samples. We compare model performance across three scenarios: without SMOTE, with traditional SMOTE, and with enhanced SMOTE, using metrics such as Accuracy, Precision, Recall, F1-Score, and Area Under the Curve (AUC). Our results show that the enhanced SMOTE significantly improves model performance, especially in recall and AUC-ROC. Notably, the CNN model with enhanced SMOTE prevailed, achieving the highest overall performance with an AUC of 0.90, an Accuracy of 0.91, a Precision of 0.89, a Recall of 0.86, and an F1- Score equal to 0.87, making it the most effective model in this study. This research highlights the potential of the enhanced SMOTE in developing robust predictive models for CVD, with broader implications for healthcare analytics.
|