Diagnose the Type II Diabetes Using Feed Forward Back Propagation Neural Networks

Diabetes mellitus is the most endocrine disease. This paper presents an optimum back propagation algorithm with analysis and its benefits for diagnosing the Type II diabetes. The Levenberg Marquadrt back propagation approach is used to analyze the convergence of back propagation with the following methods 1) selection of activation function to diagnose the diabetes mellitus 2) finding the optimal number of hidden layers to improve the performance of the system. The study has tested with 768 data, from that 268 are having diabetes. The performance of the proposed method is demonstrated by comparing the results of various activation functions in terms of accuracy and identifying the number of hidden layers in terms of minimum CPU time. It proves that 100% accuracy to be achieved when using specific combination of activation functions. Out of this hyperbolic tan sigmoid function with two hidden layer always produce accurate result with minimum number of epochs.