Prediction of Cardiovascular Disease Through Cutting-Edge Deep Learning Technologies: An Empirical Study Based on TENSORFLOW, PYTORCH and KERAS

Abstract

In healthcare system, the predictive modelling procedure for risk estimation of cardiovascular disease is extremely challenging and an inevitable task. Therefore, the attempt to clinically examine medical databases through conventional and leading-edge machine learning technologies is contemplated to be valuable, accurate and more importantly economical substitute for medical practitioners. In this research study, primarily we have exploited both individual learning algorithms and ensemble approaches including BayesNet, J48, KNN, multilayer perceptron, Naïve Bayes, random tree and random forest for prediction purposes. After analysing the performance of these classifiers, J48 attained noteworthy accuracy of 70.77% than other classifiers. We then employed new fangled techniques comprising TENSORFLOW, PYTORCH and KERAS on the same dataset acquired from Stanford online repository. The empirical results demonstrate that KERAS achieved an outstanding prediction accuracy of 80% in contrast to entire set of machine learning algorithms which were taken under investigation. Furthermore, based on the performance improvisation in prediction accuracy of cardiovascular disease, a novel prediction model was propounded after conducting performance analysis on both approaches (conventional and cutting-edge technologies). The principle objective behind this research study was the pursuit for fitting approaches that can lead to better prediction accuracy and reliability of diagnostic performance of cardiovascular disease.