A Machine Learning Approach for Predicting Heart Disease using Efficient Algorithm

Introduction, Literature Survey, Proposed Method, Experimental Results and Discussions, Conclusion

Abstract

Abstract: In India, there is a high increase in heart disease cases. It’s very important to predict such cases beforehand.  In modern medical research they have concluded that the over rate of heart disease is because of extreme levels of exercise, i.e people think that participating in jogging, bodybuilding, running, and exercises leads to good fitness but medical researchers have finalized that they may lead to the sudden death of a person at the very early age. It's difficult to predict manually whether a person is undergoing heart disease or not. So to overcome this difficulty we proposed a heart disease prediction system based on the medical history of patients. In this paper, we mainly focus on proving how best the machine learning models can be built to predict even at the worst case to overcome the early death of a patient. It is implemented considering various medical attributes. The effective algorithm called Gradient Boosting Classifier and KNN (K-nearest neighbor)  are used to classify whether a patient is diseased or not. The dataset consists of 303 instances and 14 attributes. Using this dataset in the first iteration KNN algorithm has achieved 83% of accuracy when the k value is 5, on increasing the number of instances from 303 to 918 in the second iteration we have achieved 91% accuracy for the same K value. And on experimenting with the Gradient Boosting Classifier algorithm for 918 instances we have achieved an accuracy of 93%.