Improved Accuracy for Heart Disease Diagnosis Using Machine Learning Techniques

This work primarily focuses on diagnosis of heart disease before explicit visit to the expert doctor. Machine learning based systems have been found useful in medical diagnosis applications because of their ability to learn human like expertise and to utilize acquired knowledge for diagnosis. This work is performs classification of heart disease utilizing subject’s vital parameters. Pathological laboratory results available after testing are not understood by common people and patients have to wait till they visit expert doctors for inference. In this paper, traditional methods like linear regression  to various machine learning based systems including back propagation neural network, support vector machine(SVM) and k-nearest neighbor are developed for heart diseases classification. The proposed system transforms sensor inputs to stroke stage classification. With a view to ascertain the efficacy of proposed system, performances of all methods are compared on standard Cleveland database and with similar work. Simulation results show 100 percent correct diagnosis and henceforth robustness of SVM based approaches for test data given.