Heart Failure Prediction Using Machine learning Approaches

Abstract

Heart Failure (HF) is a familiar disease that can rise to a dangerous situation in today's world. It is currently one of the most dangerous heart diseases in humans, and it seriously shortens people's lives. Heart failure can be prevented in its early stages and will increase the patient's survival if human heart disease is accurately and quickly identified. Manual methods are biased and subject to interexaminer variability when used to diagnose cardiac disease. To Predict heart failure at the correct time is difficult from the perspective of a heart specialist and surgeon. Luckily, prediction and classification models exist, which can assist the medical industry and demonstrate how to effectively use medical data. In this regard, machine learning algorithms are effective and efficient methods to identify and classify patients with heart disease and healthy individuals. According to the proposed study, we used a variety of machine learning algorithms to identify and predict human heart disease, and we used the heart disease dataset to evaluate the performance of those algorithms using various metrics, including classification accuracy, F measure, sensitivity, and specificity. Several types of machine learning algorithms are used to estimate the probability of having heart failure in a medical database. For this purpose, we used nine machine learning classifiers, including DT, LR, GBe, NB, KNN, SVM, ADB, RF, and XGB, to the final dataset before and after hyperparameter tuning. By successfully completing preprocessing, dataset standardisation, and hyperparameter tuning, we also check their accuracy on the standard heart disease dataset. Last but not least, the experimental results indicated that data standardisation and hyperparameter tuning of the machine learning classifiers significantly improved the prediction classifiers' accuracy.