• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Master's Dissertation
DOI
https://doi.org/10.11606/D.100.2022.tde-25112022-170424
Document
Author
Full name
Fernando Danilo de Melo
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Paulo, 2022
Supervisor
Committee
Rodrigues Neto, Camilo (President)
Crepaldi, Antonio Fernando
Ferreira, Fernando Fagundes
Ferreira, Leandro Augusto
Title in Portuguese
Otimização de portfólio: uma análise através de técnicas de Reinforcement Learning e Autoencoders
Keywords in Portuguese
Autoencoders
Reinforcement Learning
Aprendizado de Máquina
Criptomoedas
Dados Fundamentalistas
Dados Técnicas
Mercado de Ações
Otimização de Portfólio
Abstract in Portuguese
Com o desenvolvimento dos algoritmos de Reinforcement Learning nos últimos anos, houve um aumento no número de estudos relacionados à negociação de ativos e otimização de portfólio. Embora trabalhos com dados de análise técnica e fundamentalista ganharam notoriedade nos últimos anos, poucos incluem ambos. Outro tema pouco explorado é o impacto do uso de Autoencoders para extrair variáveis e conexões entre os dados. Buscando explorar esses pontos e entender o impacto da introdução dessas variáveis, propomos um sistema inteligente para otimizar um portfólio por meio de análises de dados técnicos e fundamentalistas, bem como as variáveis geradas utilizando Autoencoders . Avaliamos o modelo em dois mercados distintos (o mercado Norte Americano de Ações e o de Criptoativos) em mais de 10 ativos, buscando avaliar o desempenho do agente em relação a modelos tradicionais. Posteriormente, esta avaliação permitiu-nos entender o impacto dos dados dos ativos em seu desempenho e como o agente se comporta em um mercado tradicional, como o de ações, e em mercados menos regulamentados, como o de criptomoedas
Title in English
Portfolio optimization: an analysis through Reinforcement Learning techniques and Autoencoders
Keywords in English
Autoencoders
Cryptocurrencies
Fundamental Data
Machine Learning
Portfolio Optimization
Reinforcement Learning
Stock Market
Technical Data
Abstract in English
With the development of Reinforcement Learning algorithms in recent years, there has been an increase in the number of studies related to trading and portfolio optimization. Although works with technical and fundamental analysis data have gained notoriety recently, few include both of them. Another little explored subject is the impact of using Autoencoders to extract variables and connections among data. Seeking to explore these points and understand the impact of introducing these variables, we propose an intelligent system for optimizing a portfolio via analyses of technical and fundamental data as well as the variables generated through Autoencoder. We evaluated the model in ten markets (U.S. Stocks and Crypto assets) and more than 10 assets with hourly data, seeking to assess the agents performance about baselines. Subsequently, this evaluation allowed us to evaluate the impact of asset data on its performance and how the agent behaves in a more traditional market, such as stocks, and in less regulated markets, such as cryptocurrencies
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2023-05-16
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.