Analysis and evaluation of mathematical models for medical data synthesis

Title	Analysis and evaluation of mathematical models for medical data synthesis
Authors	S. F. Saifetdinov¹, T. F. Mamedova¹ ¹National Research Mordovia State University
Annotation	This study addresses the issue of data scarcity in clinical practice and evaluates methods for their artificial expansion. A comparison of the classic SMOTE algorithm with modern neural network approaches is conducted when working with ultra-small samples. Testing was conducted using a strict validation protocol (training on synthetic data, testing on real data). The results show that tabular autoencoders most accurately reproduce the internal structure of medical data, whereas SMOTE maximizes predictive accuracy but may distort the original distribution.
Keywords	data enrichment, generative adversarial networks, small samples, machine learning, data synthesis.
Citation	Saifetdinov S. F., Mamedova T. F. ''Analysis and evaluation of mathematical models for medical data synthesis'' [Electronic resource]. Mathematical modeling, numerical methods, and software systems: Collection of materials from the 12th All‑Russian Scientific Youth School‑Seminar named after E. V. Voskresensky (Saransk, July 21–24, 2026). - pp. 200-202. Available at: https://conf.svmo.ru/files/2026/papers/paper105.pdf. - Date of access: 25.09.2026.