The Lucas Tree Model in the age of AI: an agent-based reinforcement learning approach

• Ano de defesa: 2024
• Autor(a) principal: Moraes, Kauê Lopes de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Agent-based model; Aprendizado de máquina; Aprendizado por reforço; Asset pricing; Economia; Economics; Finanças; Finance; Lucas Tree Model; Machine learning; Modelo baseado em agentes; Modelo de Árvore de Lucas; Precificação; Reinforcement learning
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/12/12138/tde-14052024-164218/

This dissertation explores the integration of economic modeling and advanced machine learning techniques (reinforcement learning), with a specific focus on agent-based modeling (ABM) for the simulation of financial markets. The main goal is to develop an asset market simulation environment, crucial for deepening our understanding of the dynamics of financial markets. Utilizing the Lucas tree model, this research establishes a framework to test and validate the developed simulation techniques, given that the model has an analytical solution in some cases. The model is replicated through an agent-based approach, creating a simulated environment conducive to generating the necessary data for training artificial intelligence models. The computational project developed for this study is characterized by its flexibility, allowing the exploration of various economic scenarios and the relaxation of several traditional hypotheses in macro-finance models. This flexibility is crucial, as it enables the addressing of scenarios that are challenging to be dealt with using traditional analytical methods. The results corroborate with the effectiveness of agent-based modeling in replicating the classical economic model and in generating data for more in-depth analyses. This work not only offers new perspectives on the Lucas tree model but also establishes a basis for future research, which can expand and explore other complex facets of financial markets.