A carregar...
Publicação
Modelos bayesianos hierárquicos no planeamento de recursos humanos
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.32 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
"O planeamento de recursos humanos (PRH) tem sido tradicionalmente utilizado pelas organizações para garantir que dispõem do número certo de pessoas, no lugar e no tempo certo" (Jackson e Schuler, 1990). Do ponto de vista da metodologia estatística, uma estrutura de recursos humanos (RH) pode ser descrita como um sistema dinâmico aleatório de stocks e fluxos. Em grande parte dos modelos estocásticos de RH, propostos na literatura, o erro nas previsões não é quantificado, sendo apenas estudada a dinâmica temporal dos valores médios dos stocks e fluxos. A primeira parte da tese tem como base o trabalho desenvolvido em Vicente e Turkman (2008), onde se considerou modelos markovianos e de renovamento para modelar o sistema de RH da Faculdade de Ciências da Universidade de Lisboa. Na segunda parte deste trabalho, pretende-se caracterizar os sistemas de RH de Instituições de Ensino Superior como processos de contagem multidimensionais. O objectivo é fazer inferências sobre a dinâmica temporal da estrutura de probabilidade desta cadeia de Markov multidimensional, ao contrário do habitual estudo dos valores esperados. A cadeia é tratada como não homogénea e os fluxos, bem como os stocks, como processos aleatórios, captando a dinâmica temporal através de um modelo bayesiano hierárquico. A flexibilidade destes modelos e a capacidade do WinBUGS em ajustar uma gama considerável de modelos, permite obter predições pontuais e intervalos de credibilidade, a médio e a longo prazo, para os stocks das categorias, as promoções, os recrutamentos e as saídas, considerando vários cenários. Adicionalmente, permite analisar as políticas a ser implementadas de forma que a evolução dos RH seja a pretendida.
"Manpower planning has traditionally been used by organizations to ensure that the right person is in the right job at the right time" (Jackson e Schuler, 1990). From a statistical methodological point of view, Manpower planning can be defined as a random dynamic system of stock and flows. In general, most manpower systems proposed in the literature do not try to model the probability structure of this random dynamic system, but merely try to model the how this system in average evolves in time. Vicente and Turkman (2008) applied at standard Markov and reneval processes to model the manpower system of the Faculty of Sciences of University of Lisbon and only studied the temporal dynamics of average values of stocks and flows. This work is reported in chapters 3 and 4. The second part of this theses treats the manpower system as a dynamic multidimensional counting process of stocks. The aim is to make inference on the whole probability structure of this dynamic multidimensional counting process, which is modeled as a multivariate non-homogeneous Markov chain. Inference on this process is made using Bayesian Hierarchical models and simulation based inference. Flexibility of these methods and the corresponding software WinBUGS permit the possibility of adjusting many different variants of the proposed model, as well as study the consequences of many different man power policies that may be adopted. Proposed model also allow us to make short and long time predictions together with their credible intervals on quantities such as future stocks, promotions, recruitments as well as wastages.
"Manpower planning has traditionally been used by organizations to ensure that the right person is in the right job at the right time" (Jackson e Schuler, 1990). From a statistical methodological point of view, Manpower planning can be defined as a random dynamic system of stock and flows. In general, most manpower systems proposed in the literature do not try to model the probability structure of this random dynamic system, but merely try to model the how this system in average evolves in time. Vicente and Turkman (2008) applied at standard Markov and reneval processes to model the manpower system of the Faculty of Sciences of University of Lisbon and only studied the temporal dynamics of average values of stocks and flows. This work is reported in chapters 3 and 4. The second part of this theses treats the manpower system as a dynamic multidimensional counting process of stocks. The aim is to make inference on the whole probability structure of this dynamic multidimensional counting process, which is modeled as a multivariate non-homogeneous Markov chain. Inference on this process is made using Bayesian Hierarchical models and simulation based inference. Flexibility of these methods and the corresponding software WinBUGS permit the possibility of adjusting many different variants of the proposed model, as well as study the consequences of many different man power policies that may be adopted. Proposed model also allow us to make short and long time predictions together with their credible intervals on quantities such as future stocks, promotions, recruitments as well as wastages.
Descrição
Tese de doutoramento, Estatística e Investigação Operacional (Probabilidade e Estatística), Universidade de Lisboa, Faculdade de Ciências, 2011.
Palavras-chave
Planeamento de recursos humanos Modelos de Markov Modelos bayesianos Probabilidades e estatística Teses de doutoramento - 2011