Застосування методів машинного навчання для прогнозування стадій захворювання цукрового діабету

Abstract

This research analyzes the effectiveness of machine learning methods for predicting diabetes stages using biochemical blood test data and retinal images. The study employs deep neural networks for biochemical data analysis and convolutional neural networks for retinal image processing. A hybrid model is proposed, integrating both data types using transformer-based archІТectures. SHAP method is used for model explain abilІТy, ensuring transparency for clinical use. The results demonstrate high prediction accuracy, highlighting the potential for improving diabetes management and complication prevention.