Prediction of Heart Disease Using Fuzzy Rough Set Based Instance Selection and Machine Learning Algorithms

Abstract

In this study, instance selection was made using the fuzzy rough set based instance selection method, which is the main indicator of heart disease risk, with the finding of a certain narrowed main cardiovascular number and some other medical findings. Then, with the help of machine learning algorithms, a heart disease risk estimation model was developed over two different size and structure data sets. The heart disease dataset, formed by combining 5 different heart disease datasets, was taken from the IEEE dataport website [1]. In order to eliminate noisy instances, the 12-variable data of 1190 patients was reduced to 836 instances by fuzzy rough set based instance selection method. Qualitative variables used in the analysis are age, sex, chest pain type, resting bps, cholesterol, fasting blood sugar, resting ecg results, maximum heart rate, exercise induced angina, oldpeak, ST slope. Then, the data set was divided into two as 70% training and 30% test data sets, two-class averaged perceptron, two-class Bayes point machine, two-class logistic regression, two-class support vector machine, two-class neural network, two-class locally deep support vector machine and two-class boosted decision tree models were trained. As a result of the validity analysis carried out, the use of fuzzy rough set based instance selection method improved the prediction performance of all models. While the Two-Class Boosted Decision Tree method gave a higher accuracy than other methods, it gave an accuracy result between 89% and 93% in other methods.