A novel IRBF-RVM model for diagnosis of atrial fibrillation

•A novel IRBF-RVM model is proposed for the diagnosis of atrial fibrillation (AF).•The kernel parameter of IRBF-RVM has a much larger selectable region than RBF-RVM.•IRBF-RVM performs better than SVM in terms of rapid modeling and sparseness.•IRBF-RVM outperforms k-NN, Naive Bayes, FFNN, AdaBoost and RF, in terms of prediction accuracy.•The generalization performance of IRBF-RVM is not inferior to other powerful machine learning methods.