Mostrar registro simples

dc.contributor.advisorKleina, Mariana, 1988-pt_BR
dc.contributor.otherUniversidade Federal do Paraná. Setor de Tecnologia. Curso de Graduação em Engenharia de Produçãopt_BR
dc.creatorQueiroz, Gabriel Silva dept_BR
dc.date.accessioned2025-12-01T14:51:27Z
dc.date.available2025-12-01T14:51:27Z
dc.date.issued2021pt_BR
dc.identifier.urihttps://hdl.handle.net/1884/99548
dc.descriptionOrientador(a): Prof(a). Dr(a). Mariana Kleinapt_BR
dc.descriptionTrabalho de Conclusão de Curso (graduação) - Universidade Federal do Paraná, Setor de Tecnologia, Curso de Graduação em Engenharia de Produçãopt_BR
dc.descriptionInclui referênciaspt_BR
dc.description.abstractResumo : O número de investidores na bolsa de valores brasileira não para de crescer, entre os novos investidores há quem prefira colocar seu dinheiro em aplicações de baixo risco, ao mesmo passo que existem os que querem se aventurar nos ativos de alto risco buscando uma renda extra mensal no day trade, para esses, entre as possibilidades estão as negociações no minicontrato futuro de índice Bovespa. Para o manuseio dessas negociações, é amplamente utilizada a análise técnica, ou análise gráfica, que utiliza de diversas ferramentas para prever o movimento do mercado, entre elas, os indicadores técnicos. Nesse contexto, surgem grandes aliados para investidores, os robôs de negociação, ou expert advisors, que seguem à risca as estratégias pré-definidas pelo trader podendo fazer a leitura de até dezenas de indicadores ao mesmo tempo. O objetivo da análise técnica é, de forma geral, interpretar o passado para tentar prever o futuro, nesse sentido também trabalham as redes neurais artificiais, que se introduzem nesse trabalho buscando agregar no desempenho dos robôs de negociação e aumentar sua assertividade. Portanto, esse trabalho apresenta um estudo comparativo entre dois robôs de negociação que utilizam dos mesmos indicadores técnicos para gerar sinais de compra ou venda do minicontrato futuro de índice Bovespa. O primeiro deles utiliza os métodos clássicos da análise técnica, ou seja, interpreta os indicadores técnicos selecionados e toma a decisão de compra caso todos eles indiquem essa ação, ou venda, caso contrário. Já o outro robô utiliza a mesma leitura desses indicadores para gerar inputs para uma rede neural artificial do tipo multilayer perceptron que, após ter seus pesos sinápticos regulados por um algoritmo genético, processa os sinais dos indicadores para retornar um único sinal de compra ou venda. Por fim, apresentam-se as comparações dos resultados financeiros dos robôs no período de dez meses, sendo oito utilizados nas otimizações feitas com o algoritmo genético e dois para teste. Percebe-se a superioridade do robô que utiliza redes neurais artificiais, tendo melhor desempenho tanto no período de otimização quanto no de teste, mostrando a capacidade de generalização e reconhecimento dos padrões do mercado por parte da estrutura de rede neural multilayer perceptronpt_BR
dc.description.abstractAbstract : The number of investors on the Brazilian stock exchange continues to grow, among new investors there are those who prefer to put their money in low-risk investments, while there are those who want to venture into high-risk assets looking for extra monthly income there is no day trade, for these, between the options are listed in the mini Bovespa index futures contract. For the use of these tools, technical analysis is used, or graphical analysis, which uses various tools to predict market movements, including technical indicators. In this context, there are great allies for investors, trading processes, or Expert Advisors, who follow to the letter according to the strategy pre-defined by the trader, being able to read up to dozens of indicators at the same time. The objective of technical analysis is, in general, to interpret the past to try to predict the future, in this sense it also works as artificial neural networks, which are introduced in this work seeking to add to the performance of trading robots and increase their assertiveness. Therefore, this work presents a comparative study between two trading robots that use the same technical indicators to generate buy or sell signals for the mini Bovespa index futures contract. The first of them uses the classic methods of technical analysis, that is, it interprets the selected technical indicators and makes the decision to buy if they all indicate this action, or sell, otherwise. The other robot, on the other hand, uses the same reading of these indicators to generate inputs for an artificial neural network of the multilayer perceptron type that, after having its synaptic weights regulated by a genetic algorithm, processes the indicators' signals to return a single buy or sell signal. Finally, it is presented as comparisons of the financial results of the exercises over a period of ten months, eight being used in the optimizations made with the genetic algorithm and two for testing. It is noticed the superiority of the robot that uses artificial neural networks, having the best performance both in the optimization period and in the test, showing a generalization capacity and recognition of market patterns by the perceptron multilayer neural network structurept_BR
dc.format.extent1 recurso online : PDF.pt_BR
dc.format.mimetypeapplication/pdfpt_BR
dc.languagePortuguêspt_BR
dc.subjectRedes neurais (Computação)pt_BR
dc.subjectAlgorítmos genéticospt_BR
dc.subjectMercado financeiropt_BR
dc.subjectNegociação de futurospt_BR
dc.titleRede neural artificial aplicada à análise técnica para day trade com robôs de negociaçãopt_BR
dc.typeTCC Graduação Digitalpt_BR


Arquivos deste item

Thumbnail

Este item aparece na(s) seguinte(s) coleção(s)

Mostrar registro simples