An automatic diagnosis system based on data mining for diabetes disease

Abstract

Data mining is the process of finding meaningful data from large data stacks with the help of intelligent methods. In many areas such as biology, genetics, finance, medicine and engineering, data mining is used to obtain meaningful results from data. Knowing the characteristics of the data of the subject studied increases the success of data mining techniques. In this study, different data mining techniques, Neural Network (NN) and Decision Tree (DT), are used for diagnosis of diabete disease. The dataset for this study are obtained from Pima Indians Diabetes database. To increase classification performance, the samples in database are normalized. Normalization process for the diagnosis of diabetes is made according to the criteria of international organizations. The normalized samples are used for training and testing of NN and DT. Performances of both data mining techniques are investigated by depending on classification accuracy. The obtained results are also compared with that of the previous studies for better validation. The results of this study show that DT using normalized samples is more effective than NN using normalized samples as well as previous studies. © Sila Science.