Investigating Machine Learning Models for Effective Dataset Training in Cardiac Arrest Prediction

Abstract

Inaccuracy of data coupled with invasiveness in diagnosis of cardiac arrest is an issue of concern in clinical setting. In this study, the identification and prediction of cardiac arrest based on existing data was investigated using Machine learning (ML) algorithms. Three classic models of machine learning (Gradient Boost, Random Forest and XG Boost) models were used. Numerical variables were encoded using Label Encoder function from Scikit learn using the three models to train the data. A panda was used for data loading. After training, Gradient boosting, Random Forest and XG Boost models possess an accuracy of prediction values of 88, 89 and 85% with and an error prediction values of 23, 20 and 27, respectively. Hence fitting Gradient boosting model is the best machine learning model for training data and prediction of cardiac arrest due to its high accuracy and low error value.