Model of stochastic auctions using level market index

The following research paper is devoted to the complex topic of modeling stochastic financial markets using the example of auction markets. The presented model for market makers’ behavior on stochastic auction markets contributes practically to the field of studying portfolio optimization, risk management, market participants’ balance processes, and prediction problems via cutting-edge machine learning and statistics approaches. The reliability of the given model is proved practically with the help of modern machine learning methods of validation, namely, combinatorial splits. A client-server model for remote simulation was implemented, as well as interpreted language in C++. XGBoost, Catboost, LSTM, NN Ensemble, and H2O Auto-ML models were considered in the course of building the decision model. Hyperparameters were obtained via Optuna. Besides that, the developed model was backtested on historical data of different financial assets, starting with stocks and ending with commodity prices and foreign exchange rates. Within all models, positive Sharpe ratios have been obtained, which indicates the robustness of the model. The paper offers a valuable framework for market maker decision-making stochastic modeling, examining its pricing mechanisms and financial risk management as crucial for exchanges, funds, and other financial institutions, which makes it relevant in the context of the current dynamics of the development of financial markets and the increase in trading volumes.