SYNTHETIC DATA GENERATION FOR TRAINING A CANCER CELL SURVIVAL RATE PREDICTION MODEL FOLLOWED BY XAI ANALYSIS

The paper is devoted to the generation of synthetic data for training a model for predicting the survival rate of cancer cells under radiation and hyperthermia. Modern methods of synthetic data generation, including variational autoencoders and semantic rules based on scientific research, were investigated. The advantages and limitations of the different approaches were analyzed, as well as their applicability in the presence of small amounts of raw data. The work results in the creation of a realistic synthetic dataset, preservation of key patterns of the original data, and the development and comparison of gradient boosting and neural network models for cell survival prediction. The SHAP method was used to interpret the contribution of each parameter, which confirmed that the model was consistent with known medical patterns.