AN INTELLIGENT MODEL FOR ANALYZING BLOOD GASOMETRY BASED ON ENSEMBLE MACHINE LEARNING METHODS FOR DIAGNOSING DIABETIC KETOACIDOSIS AND SUPPORTING CLINICAL DECISION-MAKING

Abstract

Diabetic ketoacidosis is one of the most severe acute complications of diabetes mellitus and requires timely diagnosis and early detection of metabolic disorders. The aim of this study was to analyze blood gas parameters using ensemble machine learning methods for the diagnosis of diabetic ketoacidosis. Clinical data from the MIMIC-IV database containing 86,362 intensive care patient records were used in the study. Ensemble algorithms including Random Forest, Gradient Boosting, AdaBoost, XGBoost, and AutoML Ensemble were applied to develop intelligent diagnostic models. Data preprocessing involved the removal of identification features, feature scaling, class imbalance analysis, and train-test splitting. Model performance was evaluated using Accuracy, Precision, Recall, F1-score, ROC-AUC, and PR-AUC metrics. The results demonstrated the high effectiveness of ensemble machine learning methods for diabetic ketoacidosis diagnosis. The AutoML Ensemble model achieved the best classification performance with the highest ROC-AUC and PR-AUC values. The scientific novelty of the study lies in the integrated application of ensemble machine learning approaches for intelligent analysis of blood gas parameters. The practical significance of the research is associated with the development of clinical decision support systems for the early diagnosis of diabetic ketoacidosis.