Kosovych I. Modelling of epidemic spread forecasting by methods of moving cellular automata and reinforcement learning

The dissertation is devoted to modeling the spread of epidemics using mathematical models, cellular automata, and reinforcement learning methods. A detailed analysis of the main scientific works and existing mathematical models of epidemiological processes was carried out in order to identify their advantages and disadvantages. An analysis of the most common software for modeling epidemiological processes was carried out, examples of use were given. New algorithms for predicting the dynamics of infectious diseases using mobile cellular automata were developed, based on artificial intelligence methods, in particular machine learning and taking into account the spatio-temporal dynamics of the disease. To optimize the social distancing strategy, the SIHRD model using the Advantage Actor-Critic (A2C) algorithm and the SEIR model using Deep Q-Learning (DQN) were proposed, which allows adapting the agent's policy to changes in the environment. A series of experiments were conducted that confirmed the effectiveness of combining mathematical and simulation modeling for predicting epidemics and determining optimal containment measures. The process of developing, architecturally designing, testing, and deploying a web application for modeling the spread of infections has been completed, allowing users to run numerical simulations, change model parameters, and analyze results in a convenient graphical format. The findings can be used to improve pandemic response strategies and optimize public health interventions.