Prediction of heart disease outcomes using machine learning classifier

Abstract

The responsibility of heart organ is to supply blood to every part of the human body. The method of diagnose heart disease in medical hospital is extremely costly and also consume doctors time of operations. This research work applied forward, backward, and enter method for selection of variables in the logistic regression model, sensitivity, specificity, accuracy, and area under characteristic curve (AUC). The logistic regression model, at 5% level of significance with the enter method is used which denotes that the risk variables associated with heart disease gives accuracy of 87.9%. The preferred model of variable selection method used was the model from forward which has 88.6%. Also using the forward method of variables selection, the process produces 10 models with the best accuracy of 88.6%. The specificity and sensitivity of the analysis model was 91.4% and 85.6%. Also, the misclassification rate was also 11.4%, Positive predicted value is 87% and negative predicted value is 90.5%. Finally, the suitable model to predict the heart disease is from the forward method of variables selection and the positive likelihood ratio is 6 i.e the patients are 6 times likely to have the heart disease and the model has AUC value of 1.

A novel two-tier feature selection model for Alzheimer’s disease prediction

“…Regression analysis is one statistical method used in understanding the relationship between variables and can evaluate the importance of each independent parameter on the dependent parameter (Alzheimer) [19]. When single regression models are used for feature ranking in Clinical application, because of the correlation of clinical parameters and regularization behaviour, the performance of feature ranking was inaccurate.…”