Autoencoder-Based Feature Learning for Predicting Cardiovascular Disease

Abstract

Cardiovascular disease is the world's leading cause of death. Some studies have used the machine learning method to predict cardiovascular diseases based on medical records. However, due to high correlation between data in medical records, much needs to be done in the field. Here, we propose to use Autoencoder based feature learning to predict cardiovascular disease, because Autoencoder can process complex, high-dimensional datasets by doing linear and non-linear projections. Thus, we hope the autoencoder can learn about non-linear and complex connections between the medical data being used. We varied the depth of autoencoder in this paper from 3 to 7 layers, and the depth of layer was varied to several neurons at the bottleneck. The results are then input into other classifiers, such as Logistic Regression, Naive Bayes, SVM, KNN, Decision Tree, XGBoost, Random Forest, and Neural Networks. Our experiments show that use of autoencoder-based feature learning can improves the performance of the classifier by 0.75%. However, we see that the depth of the layer does not always enhance performance and needs to be defined empirically.