Grossinho, Maria do RosárioCruz, José SantosFreire, Duarte Vaz2021-12-172022-06-172021-09Freire, Duarte Vaz (2021). "Forecasting U.S. REIT index prices with artificial neural networks". Dissertação de Mestrado, Universidade de Lisboa. Instituto Superior de Economia e Gestão.http://hdl.handle.net/10400.5/22765Mestrado Bolonha em Mathematical FinanceArtificial Neural Networks are innovative mathematical models that have re- cently gathered much attention as a new tool for forecasting in economics and finance. These algorithms are characterized by being able to handle vast amounts of data and solve complex problems, without the assumption of linearity often made by traditional models. This thesis investigates the use of this Machine Learning method for forecast- ing Real Estate Investment Trusts (REIT) prices and their movement. In this experiment, we make use of a 20-year data sample related to four U.S. REIT in- dexes and other financial and macroeconomic variables. Three Neural Networks, with di↵erent architectures, were developed and we compare the results to those of traditional econometric approaches. The results show that the Neural Networks were able to outperform traditional forecasting methodologies, by registering significantly lower Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) values on their predictions. Furthermore, these models also achieved higher levels of directional accuracy, for the majority of the indexes studied.As Redes Neuronais Artificiais são modelos matemáticos inovadores usados, de forma cada vez mais frequente ao longo dos últimos anos, como ferramenta de previsão no setor económico e financeiro. Estes algoritmos caracterizam-se por ter a capacidade de lidar com grandes quantidades de dados e de resolver problemas complexos, sem a suposição de linearidade muitas vezes feita por modelos tradicionais. A presente tese investiga o uso deste método de Machine Learning na previsão de preços de fundos de investimento imobiliário (REIT). Neste estudo, usamos uma amostra de dados, correspondente a um período de 20 anos, relacionada com quatro índices REIT dos EUA e outras variáveis financeiras e macroeconómicas. Três Redes Neuronais, com arquiteturas diferentes, foram desenvolvidas e foi feita a comparaçaõ entre os resultados deste método e os resultados de modelos econométricos tradicionais. Os resultados mostram que as Redes Neuronais foram capazes de superar os métodos de previsão tradicionais, registando valores substancialmente mais baixos de Erro Quadrático Médio (RMSE) e Erro Médio Absoluto (MAE) nas suas previsões. Adicionalmente, estes modelos também alcançaram níveis mais elevados de precisão direcional, para a maioria dos índices em estudo.engReal Estate Investment TrustsArtificial Neural NetworksMachine LearningFinancial MarketsForecastingForecasting U.S. REIT index prices with artificial neural networksmaster thesis