A Framework for Cardiac Arrest Prediction via Application of Ensemble Learning Using Boosting Algorithms

Abstract

Cardiovascular diseases are a major setback for human life and a critical problem in hospitals for public health. The sedentary lifestyle of every individual lead to various heart diseases. Cardio diseases malfunction the electrical beating of the heart and result in sudden death. Thus, a novel early warning system needed to be developed for cardio patients with higher sensitivity and lower false positive rates. This work helps in the early detection of cardiac arrest by analyzing the dataset. Our work has mainly two aspects: analyzing the dataset and deriving relations among features and early diagnosis of cardiac arrest with enhanced F1-score. A number of Machine Learning algorithms were applied to conduct the experiment. The work also compares various boosting algorithms named AdaBoost, LightGBM, CatBoost and XgBoost. We used different Machine Learning algorithms in which CatBoost gave an enhanced and best prediction score 98.08%. In the case of CatBoost the predicted positive value is 98 % and Sensitivity is 97.8%.