Heart Failure Prediction Using Machine Learning and Artificial Neural Networks

This paper addresses the prediction of heart failure, a life-threatening condition that is caused by cardiovascular diseases, the number one cause of deaths worldwide. This analysis is conducted using machine learning algorithms and the backpropagation technique of artificial neural networks (ANN). A real dataset containing 12 features for 299 samples is used. First feature analysis is conducted, then three machine learning algorithms (Logistic Regression, K-Nearest Neighbors - k-NN and Naïve Bayes) were implemented along with the Multiple-Layer Perceptron (MLP) ANN using the backpropagation algorithm. For each algorithm, different values of the hyper-parameters were considered in order to study the accuracy and obtain the best results. The accuracies of the four different algorithms are 83%, 77%, 76%, and 84% for the Logistic Regression, K-NN, Naïve Bayes, and the ANN respectively. The early detection of the disease can aid in reducing the high risk imposed on patients and the algorithm proposed can become supporting tools for physicians when predicting heart failure.