Evaluation of Machine Learning Techniques in Predicting Acute Coronary Syndrome Outcome

Abstract

Data mining using machine learning techniques may aid in the development of prediction models for Acute Coronary Syndrome (ACS) patients. ACS prediction models such as TIMI and GRACE have been developed using traditional statistical techniques such as linear regression. In this paper, different machine learning techniques were evaluated to present the potential use of machine learning techniques in classification tasks as basis for future medical prediction model development. A dataset of 960 of ACS patients from the Malaysian National Cardiovascular Disease Database registry was employed and trained on three popular machine learning classifiers i.e. Naïve Bayes, Decision Tree and Neural Network to predict ACS outcome. The outcome being evaluated was whether the patient is dead or alive. An open source tool—Waikato Environment for Knowledge Analysis (WEKA) were used in executing these classification tasks. A 10-folds cross validation technique was used to evaluate the models. The performance of classifiers was presented by their accuracy rate, confusion matrix and area under the receiver operating characteristic curve (AUC). Naïve Bayes and Neural Network show generally convincing results with an average of 0.8 AUC values and 90 % accuracy rate.