Ulcerative gastroduodenal bleeding remains a serious problem in modern medicine. To determine the type of ulcerative bleeding, the Forrest classification is used worldwide, defining the tactics of treating patients with bleeding gastroduodenal ulcers. The ability to correctly classify the type of bleeding depends primarily on the experience of an endoscopist. With the development of artificial intelligence technologies, there are high expectations in improving the diagnosis and treatment of surgical diseases. This article discusses the possibility of developing an algorithm for recognizing the type of ulcerative bleeding from endoscopic images using machine learning models and integrating it into an expert medical decision support system. In the course of this study, the first Russian neural network classification was developed, which makes it possible to determine the type of ulcerative gastroduodenal bleeding with 75,56 % accuracy. The developed algorithm for recognizing the type of ulcerative bleeding is integrated into a mobile application as a tool to help in making medical decisions, which in the future will improve the quality of diagnosis and medical care for patients with gastrointestinal bleeding of ulcerative etiology.