Forecasting Diabetes Mellitus with Biometric Measurements

Authors : Emrullah ACAR, Mehmet Siraç ÖZERDEM, Veysi AKPOLAT

Pages : 28-42

View : 14 | Download : 16

Publication Date : 2011-06-01

Article Type : Research Paper

Abstract :Forecasting diabetes mellitus with biometric measurements is presented in this paper. Multilayer perceptron, Elman, ART1 Neural Networks, K-Nearest Neighbor insert ignore into journalissuearticles values(k- NN); and Support Vector Machine insert ignore into journalissuearticles values(SVM); methods were used in diabetes mellitus forecast system. The result of this study will provide alternative solutions to the medical staff in determining whether someone has diabets or not which is much easier rather than presently doing a blood test. The feedforward and feedback neural networks, K-Nearest Neighbour insert ignore into journalissuearticles values(k -NN); and Support Vector Machine insert ignore into journalissuearticles values(SVM); classifiers have been chosen for learning and testing of 768 data where 268 of them are diagnosed with diabetes. For forecasting system, 8 different biometric measurements were used. These parameters are; number of times pregnant, plasma glucose concentration, blood pressure, triceps skin fold thickness, serum insulin, body mass index, diabetes pedigree function and age. Different structures of networks were tested and the results are compared in terms of testing performance for each network model. The main purpose of this study is to forecast whether someone has diabetes or not. Finally, the best performance was observed as 87.06% in the LS-SVM model structure
Keywords : Forecast, Diabetes Mellitüs, Artificial Neural Networks ANNs, K Nearest Neighbor k NN, Support Vector Machine