Diabetes symptoms appear as high blood sugar and low insulin levels. It can be concluded that this disease is the result of changes in the function and behavior of insulin in the body. People with diabetes suffer from high glucose levels in the body. If the disease is not diagnosed timely, it will cause many complications. In this thesis, the design of an intelligent precision system based on data mining was discussed in order to diagnose diabetes. The data set used for this purpose is Pima Indian diabetes dataset. In this work, as an aspect of innovation, the combination of artificial bee colony algorithm and probabilistic neural network has been employed. Among the advantages of the probabilistic neural network, is that this network operates faster than other networks and is not sensitive to outlier data. This four-layer neural network contains a specific parameter called the Spread parameter in its pattern layer, which determines the similarity of the input vector with the other vectors in the pattern layer. The optimal adjustment of this parameter has a significant impact on the accuracy of the classifier, because if the value of this parameter is too low or too high, the neural network will not be trained properly and the classification accuracy will decrease. Therefore, in this research, the bee algorithm deployed to optimally adjust the spread parameter, and finally we were able to achieve an accuracy of 97.6%.
