Estimating Trading Portfolio Losses Using Deep Generative Models

The financial sector has been undergoing a constant process of evolution for a long time. During this period, researchers came to realize the deep relationship between mathematical principles and price fluctuations for various financial assets. This led to the conclusion that deep knowledge in mathematics and statistics can significantly contribute to successful trading in financial markets. From that moment, active efforts began to develop risk assessment methods, starting with the analysis of individual stocks and gradually moving on to a comprehensive study of the risks of investment portfolios. This ensured the formation of the basic principles of modeling and risk management in the financial sector. Currently, there are a variety of models on the market designed to evaluate and predict asset prices. Among them, the widely used ARIMA and GARCH models stand out. Nevertheless, with the development of technology, it has become obvious that the use of the latest machine learning methods greatly simplifies the forecasting of price dynamics. Moreover, the capabilities of generative models open new perspectives in the analysis and forecasting of financial markets, providing unique tools for identifying patterns and trends.