Machine Learning Methods Comparison and Prediction for Heart Failure

Abstract:
In this paper, we use different machine learning algorithms to learn and predict the binary outcome of whether one would have heart failure. The results from this study show that the XGBoost model outperforms all other machine learning algorithms with accuracy near 0.852, recall of 0.918, and precision of 0.818. Of all the attributes, the slope of the ST curve in the electrocardiogram turns out to be the most important variable. Future work can develop on more industry-specific models and collect more data on electrocardiograms to feed in the existing model.