| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 662.94 KB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
A Internet está a revolucionar o conceito de sistema de apoio a decisão, permitindo o acesso ubíquo a modelos e métodos de base matemática para problemas de gestão operacional. A interface, através de um Web browser, deve ser naturalmente simples e intuitiva, podendo o armazenamento de dados e a complexidade computacional dos métodos ser relegados para um servidor remoto. Deverá restar ao utilizador introduzir, de forma facilitada, os elementos necessários de consulta, e eventuais informações subjectivas, tais como requisitos, preferências ou estimativas circunstanciais. Neste trabalho e demonstrada uma aplicação Web para previsão de processos de procura irregularmente espaçada no tempo, e inferência de decisões optimais. Na sua criação foram tidos em conta critérios como facilidade de utilização, interface intuitivo e possibilidade de correr numa grande variedade de sistemas operativos. Foram implementadas três características particulares, para auxiliar a decisão: a apresentação das previsões através de funções densidade de probabilidade, obtidas por soma ponderada de funções kernel; a possibilidade de combinação dessas previsões com previsões subjectivas, introduzidas por estimativas intervalares; e a inferência de recomendações (soluções prescritivas) optimais face a funções de custo realistas.
The Internet is revolutionizing the concept of decision support system, allowing ubiquitous access to mathematical models and methods for operational management problems. The interface, through a Web browser, should be naturally simple and intuitive, with the possibility of relegating data storage and computational complexity issues to a remote server. The user would only need to insert, in a facilitated way, the necessary elements and possible subjective information, such as requirements, preferences and estimates. In this work we describe a Web application for demand processes forecasting and optimal decision inference. Criteria like ease of use, intuitive interface and ability to run in a wide variety of operating systems were taken into account for its creation. Apart from irregular demand time series forecasting, other three particular features were implemented to aid decision-making: the production of forecasts through probability density functions, obtained by weighted sum of kernel functions; the possibility of combining those forecasts with subjective guesses, specified by interval estimates; and the inference of optimal recommendations (prescriptive solutions) with respect to realistic cost functions.
The Internet is revolutionizing the concept of decision support system, allowing ubiquitous access to mathematical models and methods for operational management problems. The interface, through a Web browser, should be naturally simple and intuitive, with the possibility of relegating data storage and computational complexity issues to a remote server. The user would only need to insert, in a facilitated way, the necessary elements and possible subjective information, such as requirements, preferences and estimates. In this work we describe a Web application for demand processes forecasting and optimal decision inference. Criteria like ease of use, intuitive interface and ability to run in a wide variety of operating systems were taken into account for its creation. Apart from irregular demand time series forecasting, other three particular features were implemented to aid decision-making: the production of forecasts through probability density functions, obtained by weighted sum of kernel functions; the possibility of combining those forecasts with subjective guesses, specified by interval estimates; and the inference of optimal recommendations (prescriptive solutions) with respect to realistic cost functions.
Descrição
Tese de mestrado, Investigação Operacional, Universidade de Lisboa, Faculdade de Ciências, 2009
Palavras-chave
Previsão Sistemas de apoio à decisão Aplicações web Séries temporais irregularmente espaçadas Teses de mestrado - 2009
