A Hybrid Machine Learning Approach to Cardiovascular Disease Prediction

Abstract

The healthcare sector is crowded with numerous amount of patients’ data on daily basis. If this data can be mined, it could benefit the society to an extreme level. Many researches have been put forward to use such enormous data efficiently for the prediction of catastrophic diseases that benefits human kind. Cardiovascular diseases are one such kind of catastrophic diseases that if diagnosed early and accurately can save a person’s life. Thus, data mining has emerged out as a boon to healthcare sector. Numerous techniques have been researched for prediction of diseases like Support Vector Machine, Random forest, Neural Network etc. and they are scaled on accuracy parameter for efficient diagnosis of diseases. But it has been analyzed that single algorithm results in less accurate results while its ensemble with other algorithm results in higher accuracy and better prediction. In this paper, ensemble of SVM and RF is done to enhance the prediction accuracy of the system. The experiment is implemented on real data taken from Fortis Escorts, Faridabad, India along with data from UCI repository. Ensemble of different algorithms is performed like SVM and Neural Network, Neural Network and Random Forest, SVM and Random Forest and it has been found that SVM and RF results in higher accuracy as compared to ensemble of other techniques. Accuracy of 90.1% has been achieved using SVM and RF hybrid approach.