Kernel FCM-Based ANFIS Approach to Heart Disease Prediction

Abstract

Heart disease is the main reason for deaths currently in the world. This disease not only affects the old people but also middle-aged and young people. Therefore, the early and precise detection of this disease using intelligent techniques has gained a lot of importance. The goal of this paper is to introduce a diagnostic tool for the detection of heart disease using kernel-based fuzzy C-means clustering, FCM-based adaptive neuro-fuzzy inference system (ANFIS). In the conventional FCM clustering, Euclidean distance is used to compute the distance measure between data points during the clustering process. In kernel-based FCM (KFCM), kernel functions are used to compute this distance measure that enables mapping dataset to high-dimensional space in which data is clearly separable. This generalization helps to make experimental input–output dataset better and distinctly separable leading to more precise data partitions and therefore, more accurate cluster centers. Therefore, these cluster centers when used in fuzzy rule base induction can be used to construct a more precise rule base for the ANFIS which would increase the prediction performance of the system in the analysis of heart disease. For the evaluation of the proposed system, we employed the Cleveland Heart Disease data from the UCI machine learning repository.