Heart Disease Prediction Integrating UMAP and XGBoost

Abstract

The healthcare industry is flooded with the plethora of data about the patients which is supplemented each day in the form of medical records. Researchers have been putting in various efforts to bring this data into usage for the prediction of various diseases. Prediction of heart diseases is one such area. Data mining algorithms have been at the centre of improving the prediction of accuracy of heart diseases. But it has been found that these algorithms are not using adequate set of attributes for prediction that sometimes may lead to wrong predictions. The aim of this paper is to deploy the right set of algorithms to accurately predict the heart diseases and help both the patient and the doctor. The paper thrives to put UMAP and XGBoost techniques in this regard and exploit the advantages of both techniques. UMAP helps in dimensionality reduction without loss of useful data while XGBoost uses parallelization for tree construction reducing the time required to get the results. The experiment is carried on real data taken from Fortis Escorts, Faridabad, India. The results are compared with existing techniques such as Naïve Bayes, Decision Tree model, Logistic Regression model and Support Vector Machine (SVM) model based on various parameters such as accuracy, recall and precision. Remarkable accuracy of 94.59%, recall of 87.87, precision of 100 has been achieved.