Análise de estratégias e métodos para predição de fundos de investimentos imobiliários
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidade Federal de Viçosa
Abstract
Este trabalho aborda a previsão da variação do preço das cotas de Fundos de Investi- mento Imobiliário (FIIs), uma questão de relevância crescente para investidores e gestores de portfólio. Diante da complexidade desse desafio e da necessidade de tomar decisões informadas no mercado imobiliário, esta pesquisa explora estratégias inovadoras de mo- delagem, contrastando métodos tradicionais de séries temporais com abordagens de apren- dizado de máquina. Um aspecto inovador foi o enriquecimento dos dados com informações adicionais sobre os imóveis subjacentes aos FIIs, buscando entender como características específicas influenciam sua valorização. Os resultados demonstraram que, em muitos ca- sos, os modelos de aprendizado de máquina superaram as técnicas tradicionais de análise de séries temporais, especialmente na previsão de tendências de médio prazo. Além disso, este estudo investigou a viabilidade de utilizar um classificador para determinar a estraté- gia mais eficiente na predição, considerando os resultados obtidos com os modelos ARIMA e XGBoost. A análise revelou que diferentes fundos exibem comportamentos distintos ao longo do tempo, e a escolha do algoritmo adequado pode variar conforme as características individuais de cada fundo. Palavras-chave: Fundo de Investimento Imobiliário, Aprendizado de Máquina, Séries Temporais
This work focuses on predicting the variation in the price of Real Estate Investment Trusts (REITs), an increasingly relevant issue for investors and portfolio managers. Given the complexity of this challenge and the need to make informed decisions in the real estate market, this research explores innovative modeling strategies, contrasting traditional time series methods with machine learning approaches. An innovative aspect was the enrich- ment of data with additional information about the underlying properties of the REITs, seeking to understand how specific characteristics influence their appreciation. The results demonstrated that, in many cases, machine learning models outperformed traditional time series analysis techniques, especially in forecasting medium-term trends. Additionally, this study investigated the feasibility of using a classifier to determine the most efficient strat- egy in prediction, considering the results obtained with ARIMA and XGBoost models. The analysis revealed that different funds exhibit distinct behaviors over time, and the choice of the appropriate algorithm may vary depending on the individual characteristics of each fund. Keywords: Real Estate Investment Fund, Machine Learning, Time Series
This work focuses on predicting the variation in the price of Real Estate Investment Trusts (REITs), an increasingly relevant issue for investors and portfolio managers. Given the complexity of this challenge and the need to make informed decisions in the real estate market, this research explores innovative modeling strategies, contrasting traditional time series methods with machine learning approaches. An innovative aspect was the enrich- ment of data with additional information about the underlying properties of the REITs, seeking to understand how specific characteristics influence their appreciation. The results demonstrated that, in many cases, machine learning models outperformed traditional time series analysis techniques, especially in forecasting medium-term trends. Additionally, this study investigated the feasibility of using a classifier to determine the most efficient strat- egy in prediction, considering the results obtained with ARIMA and XGBoost models. The analysis revealed that different funds exhibit distinct behaviors over time, and the choice of the appropriate algorithm may vary depending on the individual characteristics of each fund. Keywords: Real Estate Investment Fund, Machine Learning, Time Series
Description
Citation
DINIZ, Henrique Penna Barbosa. Análise de estratégias e métodos para predição de fundos de investimentos imobiliários. 2024. 79 f. Dissertação (Mestrado em Ciência da Computação) - Universidade Federal de Viçosa, Viçosa. 2024.
