Improving the efficiency of classification of gastrointestinal disease phenotypes using data processing methods

The article is devoted to the problem of improving the accuracy of classification of phenotypes of diseases of the gastrointestinal tract. In the course of preliminary research, the CatBoost model was identified, which demonstrates the best results in this task, achievingan accuracy of 92.85 % in the training sample and 79.31 % in the test sample [1]. The main goal of this work is to improve the accuracy of predictions of the model solving the problem of classification of diseases by studying the characteristics of the original data set. Data processing methods aimed at increasing the quality of model predictions, including the search and removal of abnormal data in the sample, are proposed. The issue of choosing the criterion of "abnormality" for specific data is also considered. In addition, the article discusses methods of dealing with the problem of class imbalance, which is an important aspect for improving the overall efficiency of classification.