Evaluation of classifier models for the detection of diabetes disease

Abstract

Diabetes is one of the major diseases which is commonly found among all age groups and people of different origins. Diabetes is a disease which may lead to the failure of different organs, and cause high risk of blindness, kidney failure, heart disease and problems in the nervous system. Data mining algorithms could be used as alternative way for diagnosis by discovering patterns from the history of patient data and also by capturing the experience of experts. In this research different classifier models ware designed and implemented to predict type one and type two diabetes diseases. Different performances measure was evaluated to identify an optimal classifier accuracy models. The classifiers used in the experimental approaches are Decision Tree (C4.5), Support Vector Machines (SVM), and Artificial Neural Networks (ANN). The optimal models identification was done using performance evaluation matrices include accuracy (Acc), specificity (Spe.), sensitivity (Sen) and precision (Pre.). The models was tested with Pima Indian Heritage diabetes database from University California Irvine (UCI) Machine learning repository and Virginia Commonwealth University (VCU) database collected across 139 hospitals in United states of America (USA).

Keywords: Data Mining Classification, Decision tree, Artificial neural network, Support vector machine