Optimized Random Forest Algorithm with Parameter Tuning for Predicting Heart Disease

Abstract

The medical industry produces a large volume of data. Using this data, a disease can be detected, predicted and also cured. Heart diseases are one of the complex diseases and many people are suffering from this disease across the world. However, if the disease is detected early on, the mortality risk may be minimized. It is necessary to identify whether or not a person is at risk of heart disease in advance to reduce number of deaths occurring globally. This is a field in which researchers are working to help the medical practitioners to take decision regarding prediction of heart disease. To improve the accuracy for heart disease prediction is the primary goal of this research work. The cleveland dataset is used for our work which is obtained from the University of California, Irvine (UCI) Machine Learning Repository. There are 303 samples in the dataset, with 13 input features and one output feature. Here the train test ratio used is 90:10. Various classification algorithms are applied on the datasets to identify the most efficient algorithm but random forest (RF) algorithm has shown maximum accuracy in prediction. To further improve the accuracy of prediction system parameter tuning is done on the random forest algorithm.