Pessimistic multi-granulation rough set-based classification for heart valve disease diagnosis

Abstract

The primary contribution of this study relies on proposing a new method, which can detect heart diseases in respective heart valve data. In this work, supervised quick reduct feature selection algorithm is applied for selecting important features from heart valve data. The classification method is applied only for relevant features selected using supervised quick reduct from heart valve data. In this paper, a new classification approach based on pessimistic multi-granulation rough sets (PMGRS) is applied for heart valve disease diagnosis. In multi-granulation rough sets, set approximations are well-defined by multiple equivalence relations on the universe, leading to an effective model for classification. This is confirmed by experimental evaluation, which shows excellent classification performance and also demonstrates that the proposed approach is superior to other benchmark classification algorithms including naive Bayes, multi-layer perceptron (MLP), and J48 and decision table classifiers.