| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 1.99 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Levar a cabo decisões de investimento acertadas é extremamente importante para um investidor ou gestor de investimentos. Deste modo os investigadores têm-se debruçado com maior foco, na última década, na realização de estudos sobre a volatilidade dos instrumentos financeiros.
Estes, desempenham um papel de extrema relevância na definição da estratégia de negociação ou até na determinação do momento apropriado para negociar, mas sobretudo na caracterização e análise de risco de um determinado valor mobiliário. À medida que a tecnologia vai avançando, surgem ferramentas e modelos que auxiliam os analistas. É nesse sentido que a motivação para esta investigação nasce, cujo intuito é a abordagem de diferentes estratégias, aplicando o modelo heterocedástico GARCH e o algoritmo de Machine Learning XGBOOST, a series temporais relacionadas com Exchange Traded Funds (ETF’s) de mercados emergentes e desenvolvidos, de modo a prever a volatilidade destes ativos financeiros, durante o início da invasão russa ao território da Ucrânia, com recurso à linguagem de programação Python para uma amostra de
dados cujo horizonte temporal é de 01 de janeiro de 2012 a 24 de abril de 2022 e fonte é o Yahoo Finance. Os resultados apresentados, com especial foco no algoritmo XGBOOST, sugerem que a utilização de algoritmos de Machine Learning permite alcançar métricas de erro substancialmente inferiores, comparativamente ao modelo GARCH. Paralelamente é possível constatar igualmente que a volatilidade nos mercados emergentes é superior à apresentada nos mercados desenvolvidos, resultado do reduzido desenvolvimento das empresas, insegurança dos investidores e o peso que os eventos adversos, com origem em países desenvolvidos, têm
sobre estes mercados.
Making the right investment decisions is extremely important for an investor or investment manager. Therefore, researchers have focused more intensively over the past decade on conduction studies on the volatility of financial instruments. These play an extremely relevant role in defining trading strategy or even determining the appropriate timing for trading, but above all in characterizing and analyzing the risk of a particular security. As technology advances, tools and models emerge to assist analysts. It is in this sense that the motivation for this research arises, with the aim of approaching different strategies, applying the GARCH heteroskedastic model and the XGBOOST Machine Learning algorithm to time series related to ETF’s from emerging and developed markets, in order to predict the volatility of these financial assets during the onset of the Russian invasion of Ukrainian territory, using the Python programming language for a data sample with a time horizon from January 1, 2012, to April 24, 2022 sourced from Yahoo Finance. The results presented with special focus on the XGBOOST algorithm, suggest that the use of machine learning algorithms allow us to achieve substantially lower error metrics compared to the GARCH model. At the same time, it is also possible to observe that volatility in emerging markets is higher than that in developed markets, because of reduced company development, investor insecurity, and the impact that adverse events, originating from developed countries, have on these markets.
Making the right investment decisions is extremely important for an investor or investment manager. Therefore, researchers have focused more intensively over the past decade on conduction studies on the volatility of financial instruments. These play an extremely relevant role in defining trading strategy or even determining the appropriate timing for trading, but above all in characterizing and analyzing the risk of a particular security. As technology advances, tools and models emerge to assist analysts. It is in this sense that the motivation for this research arises, with the aim of approaching different strategies, applying the GARCH heteroskedastic model and the XGBOOST Machine Learning algorithm to time series related to ETF’s from emerging and developed markets, in order to predict the volatility of these financial assets during the onset of the Russian invasion of Ukrainian territory, using the Python programming language for a data sample with a time horizon from January 1, 2012, to April 24, 2022 sourced from Yahoo Finance. The results presented with special focus on the XGBOOST algorithm, suggest that the use of machine learning algorithms allow us to achieve substantially lower error metrics compared to the GARCH model. At the same time, it is also possible to observe that volatility in emerging markets is higher than that in developed markets, because of reduced company development, investor insecurity, and the impact that adverse events, originating from developed countries, have on these markets.
Description
Mestrado em Análise Financeira
Keywords
ETF’s Mercado Previsão Volatilidade Forecasting Market Volatility
Citation
Fernandes, B. M. L. (2024) A certeza da incerteza no mercado de ETF’s – Análise comparativa do modelo GARCH e algoritmo de inteligência artificial XGBOOST. [Dissertação de mestrado, Instituto Superior de Contabilidade e Administração de Lisboa]. Repositório Científico do Instituto Politécnico de Lisboa. http://hdl.handle.net/10400.21/17916