The Development of a Genetic Method for Predicting the Incidence of Diabetes

In this study, we resolve the problem of predicting the incidence of diabetes on the basis of clinical data on human health by machine learning. A genetic method has been developed to optimize the neural network weights for solving the problem of predicting the risk of diabetes. A method of selecting the individuals for parental pairs based on their relatedness and two mutation methods of chromosome genes are proposed. The main difference between the created methods and the existing ones is that when the mutations the values of the selected genes are calculated on the basis of their average values and then the value is added, subtracted or reset. The obtained results allow to offer effective methods for improving the quality of predicting the presence of diabetes, which will improve the process of diagnosis and treatment of the disease.