Heart Failure Risk Stratification in an IoT Healthcare Platform using Ensemble Learning

This paper presents a predictive model founded on Machine Learning (ML) techniques to identify patients at Heart Failure (HF) risk. This model is an ensemble learning approach, a modified stacking technique, that uses three ML models leveraging clinical and echocardiographic features and then a meta-model to combine the predictions of these three models. We assessed the model on a real dataset, and our results suggest that it is better than the chosen baselines in the stratification of patients at HF risk. Specifically, we obtained high sensitivity (91%) and good accuracy (78%). To better study the value and potential of our model, we also contrasted its results with those obtained by using selected baselines. The preliminary results indicate that our predictive model outperforms these baselines that flatly consider features, i.e., not grouping features in clinical and echocardiographic features.