In Today's world there is an increase in the prevalence of diabetes mellitus and therefore the disease is recognising as a major global public health problem.medical data mining extracts hidden patterns from medical data. This is to design system for diabetes prediction.. The soft computing technique is most useful and powerful technique used for diagnosis purpose. The proposed system a novel approach for diagnosis of diabetes which has two stages to predict the diabetes status. Initial Stage we are using Gaussian kernel function which help to distribution of data and second stage adopt two computational intelligence and knowledge engineering technique such as fuzzy logic and neural network. The benefit applying these is that accuracy of prediction rate will be higher than most of the suggested system for predicting the occurrence of diabetes mellitus. The dataset used for the Experimental is based on Pima Indian Dataset from University of California.