A Parallel Ensemble Fuzzy Classifier for Diabetes Diagnosis

Diabetes is one of the deadliest disease on the planet. It isn't just an ailment yet additionally a maker of various types of maladies like heart assault, blurred vision, nephropathy and dyspnea. When decision-making process by traditional machine learning methods for a patient is made, it often face the following challenges: (1) some uncertain factors exist in the patient or the decision-making process which often result in misdiagnosis: (2) the decision-making process with traditional machine learning methods are block-box which are not interpretable. In this paper, a parallel-based fuzzy partition and fuzzy weighted ensemble TSK (Takagi-Sugeno-Kang) fuzzy classifier called FP-TSK-FW is proposed for diabetes diagnosis by utilizing its strong uncertainty-handling capability and interpretability so as to achieve promising classification performance. In FP-TSK-FW, the training dataset firstly is partitioned into several subsets by fuzzy clustering algorithm FCM on certain attributes, each interpretable TSK fuzzy subclassifier on each training subset can be quickly built in parallel, and with different structures. Finally, the final prediction of FP-TSK-FW is realized by fuzzy weighted for the results of each classifier. The experimental results on Pima Indians Diabetes dataset indicate the effectiveness of the proposed methods in the sense of both enhanced classification performance and interpretability.