Mostrar registro simples

dc.contributor.advisorMatioli, Luiz Carlos, 1961-pt_BR
dc.contributor.otherUniversidade Federal do Paraná. Setor de Tecnologia. Programa de Pós-Graduação em Métodos Numéricos em Engenhariapt_BR
dc.creatorAndrade, Juarez Siedleckipt_BR
dc.date.accessioned2024-05-16T14:05:52Z
dc.date.available2024-05-16T14:05:52Z
dc.date.issued2006pt_BR
dc.identifier.urihttps://hdl.handle.net/1884/24177
dc.descriptionOrientador: Prof. Dr. Luiz Carlos Matiolipt_BR
dc.descriptionDissertação (mestrado) - Universidade Federal do Paraná, Setor de Ciências Exatas e Setor de Tecnologia, Programa de Pós-Graduação em Métodos Numéricos em Engenharia. Defesa: Curitiba, 06/10/2006pt_BR
dc.descriptionBibliografia: fls. 66-70pt_BR
dc.descriptionÁrea de concentração: Programação matemáticapt_BR
dc.description.abstractResumo: Neste trabalho, sugere-se uma aplicação para o Método de Lagrangeano Aumentado com Penalidade Quadrática ao problema de compor carteiras de investimentos, conforme formulação proposta por Markowitz [1]. Os Métodos de Lagrangeano Aumentado partem normalmente de funções de penalização : R R, estritamente convexas e crescentes, que são combinadas com multiplicadores de Lagrange para compor termos de penalização com os formatos:(y, ) Î R x R++ p(y, u) = (y) e (y, ) Î R x R++ p(y, u) = (y). Aqui se utiliza no algoritmo de Lagrangeano Aumentado uma função de penalização , definida por y Î R (y) = ½y2 + y, sendo estritamente convexa, porém não crescente em todo o seu domínio. Neste caso, em que as penalidades são quadráticas, os multiplicadores gerados pelo algoritmo de Lagrangeano Aumentado podem ser negativos. Este problema é contornado aumentando-se o parâmetro de penalidade, conforme mostrado por Matioli [2]. É muito importante salientar que a composição de carteiras de nvestimentos com a aplicação do Método do Lagrangeano Aumentado com Penalidade Quadrática (LAPQ), aqui proposta, consiste de uma ferramenta de avaliação para o investidor que baseada em uma série histórica encontra uma solução ótima. Portanto, não visa min ½ xT Q x s. a x = 1 x ³ 0 onde Q Î Rnxn, x Î Rn 12 substituir em momento algum o conhecimento e a capacidade de análise do investidor ou do agente responsável pelo investimento, mas sim contribuir no processo de tomada de decisão.pt_BR
dc.description.abstractAbstract: In this work, we propose an application for the Augmented Lagrangian Method with Quadratic Penalty for the problem of composed portifolio selection, conform as the formulation proposed by Markowitz [1]. The Augmented Lagrangian Method normally starts with penalization functions : R R, strictly convex and increasing, which are combined with Lagrange multipliers of to compose terms of penalization as following: (y, ) Î R x R++ p(y, u) = (y) e (y, ) Î R x R++ p(y, u) = (y). Here, in the Augmented Lagrangian algorithm, we use the penalization function , defined by y Î R (y) = ½ y2 + y, strictly convex, however it is not increasing in all its domain. In this case, where the penalties are quadratic, the multipliers generated by the algorithm of Augmented Lagrangean can be negative. This problem is solved by increasing the penalty parameter, as shown by Matioli [2]. It is very important to point out that the composition of portfolio selection with the application of the Augmented Lagrangian Method with Quadratic Penalty (LAPQ), considered here, consists of a tool of evaluation for the investor which’s based on a historical series finds an optimal solution. Therefore, this approach does not aim at any moment to substitute the knowledge and the capacity of analysis of either the investor or min ½ xT Q x s. a x = 1 x ³ 0 where, Q Î Rnxn, x Î Rn 13 the responsible agent for the investment, but to contribute in the process of decision taking.pt_BR
dc.format.extent70f. : il., grafs., tabs.pt_BR
dc.format.mimetypeapplication/pdfpt_BR
dc.languagePortuguêspt_BR
dc.relationDisponível em formato digitalpt_BR
dc.subjectLagrange, Funções dept_BR
dc.subjectInvestimentos - Análisept_BR
dc.subjectMercado de ações - Previsãopt_BR
dc.subjectProgramaçao quadraticapt_BR
dc.subjectAnálise numéricapt_BR
dc.titleLagrangeano aumentado aplicado a problemas de carteiras de investimentospt_BR
dc.typeDissertaçãopt_BR


Arquivos deste item

Thumbnail

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

Mostrar registro simples