Multiclass Diagnosis Model for Heart Disease using PSO based SVM

Abstract

One of the most significant domains of Machine learning is in the Healthcare Industry which helps the medical professionals in the automation of medical diagnosis process and in the development of a disease prediction system that is highly powerful in reducing the patient mortality rate. The count of people dying every year from heart disease is increasing drastically. The multiclass model for diagnosing heart disease has been proposed using PSO based SVM. It classifies the heart disease into 5 classes namely healthy, low-risk, medium-risk, high-risk and danger. The severity of disease increases from healthy to danger. Principal Component Analysis has been used as a dimensionality reduction step to choose the subset of attributes that best reflects the original heart dataset. Support vector machine is a promising supervised method that classifies data by functional hyperplane which separates two classes from each other. The accuracy of SVM was enhanced by global stochastic optimization technique called Particle swarm optimization. The proposed algorithm is implemented for both 2-class and 5-class problems. The performance of multiclass model has been estimated on the basis of accuracy, recall, precision and F-measure. The results indicate that the attained classification accuracy is very promising as compared to the other existing algorithms. The proposed approach can be successfully used for determining the severity level of heart disease.