Development and research of a genetic method for medical diagnosis of diabetes mellitus

• Дата: 2021
• Автори: Федорченко, Є. М., Олійник, А. О., Корнієнко, С. К., Степаненко, О. О., Саман, М. В.
• Формат: Стаття
• Мова: Ukrainian
• Опубліковано: Інститут проблем реєстрації інформації НАН України 2021
• Теми: генетичний алгоритм; класифікація; нейронна мережа; прогнозування; діабет; PSO; genetic algorithm; classification; neural network; prediction; diabetes; Keras; Python
• Онлайн доступ: http://drsp.ipri.kiev.ua/article/view/239216
• Назва журналу: Data Recording, Storage & Processing

The research is devoted to Software Engineering which connected with Parametric Synthesis of Neural Networks based on the Evolution Approach and its use in Diabetes Diagnosing. As a result of the research, the Intellectual Support System of Taking Decision for Diabetes Diagnosing was developed based on the Machine Learning Models. The Model of Parametric Optimization for Neural Networks was implemented by applying Genetic Algorithm and Particle Swarm Method. The Modified Genetic Method of Optimization Parameters for Neural Networks was developed for solving the challenge of forecasting the risk of diabetes appearance. Modification of Simple Genetic Algorithm, which has been implemented in this project, gives the opportunity to speed up the selection of parameters for learning of Neural Networks and to raise the Resultant Indicator of Precision in comparison with the basic version of simple genetic algorithm. All those were reached by modification of Operator Mutation and the modified approach to the choice of individuals for crossing. The developed model is designed for use in Medical Service and allows define with a certain precision the risk of having diabetes  in accordance with the clinical indicators of patient health status. The result of applying this model is decreasing the probability of a doctor's mistake, increasing certainty in a doctor’s diagnosis and more numbers of saved lives because of correct and timely diagnosis. Tabl.: 16. Fig.: 17. Refs: 18 titles.