This paper presents a performance of various computational approaches for diagnosis of diabetes to predict the levels of diabetes risk with better accuracy.The proposed tool comprises of all the computational techniques for first level diagnosis of diabetes. Rule based approach is applied for the results obtained from the first level diagnosis to categorize the risk level of patients. The significance of this paper is the data used in the training phase which is obtained from huge number patient’s data. Based on the observation of patient details some of the influenced parameter for diabetes diagnosis was identified. The morality of the diagnosis of diabetes is also considered to reduce the percentage of inaccurate prediction. The accuracy of the prediction rate for diagnosing diabetes was found to be 95%. © 2016, Institute of Integrative Omics and Applied Biotechnology. All rights reserved.