Loading...
Publication
Uma introdução às redes Bayesianas
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 1.46 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
As redes bayesianas são modelos de representação da realidade flexíveis por lidarem
com incerteza e vários tipos de relações entre variáveis. Estas são compostas por uma parte
gráfica, representada por grafos acíclicos direcionados, onde cada vértice representa uma
variável aleatória, e uma parte probabilística, referente às distribuições associadas a cada um
dos vértices. A estrutura do grafo define uma forma de fatorização da probabilidade conjunta
das variáveis. Estas redes são úteis na inferência probabilística, facilitando o trabalho dos
especialistas, permitindo diagnosticar, prever e realizar raciocínio intercausal.
Esta dissertação é constituída por sete capítulos. No primeiro é descrita a evolução dos
modelos do conhecimento humano. No segundo são apresentados conceitos e definições
necessários para a construção e utilização das redes bayesianas. O terceiro apresenta os
métodos de inferência nestas redes, o quarto as técnicas de aprendizagem, e o quinto a
análise de conflitos. No sexto são apresentados alguns comandos do programa R úteis na
aplicação dos conceitos apresentados e no sétimo são apresentadas as considerações finais.
Bayesian networks are reality representing models that are flexible given that they handle uncertainty and different kinds of relationships among variables. These have a graphical aspect, given their directed acyclic graph structure, where each node represents a variable, and a probabilistic aspect, corresponding to the probabilistic distributions associated with each node. The graph structure defines a possible factorization of the joint probability of the variables. These networks are useful for probabilistic inference, helping experts in their work to perform tasks such as diagnosis, forecasts and inter-causal reasoning. This dissertation is composed by seven chapters. The first consists of a description of the evolution of the human knowledge models. In the second, concepts and definitions necessary for the construction and use of these networks are introduced. The third presents the inference methods, the fourth presents different learning approaches and the fifth presents the conflict analysis. The sixth chapter explores some commands of R language related to bayesian concepts and inference and on seventh are presented some final considerations.
Bayesian networks are reality representing models that are flexible given that they handle uncertainty and different kinds of relationships among variables. These have a graphical aspect, given their directed acyclic graph structure, where each node represents a variable, and a probabilistic aspect, corresponding to the probabilistic distributions associated with each node. The graph structure defines a possible factorization of the joint probability of the variables. These networks are useful for probabilistic inference, helping experts in their work to perform tasks such as diagnosis, forecasts and inter-causal reasoning. This dissertation is composed by seven chapters. The first consists of a description of the evolution of the human knowledge models. In the second, concepts and definitions necessary for the construction and use of these networks are introduced. The third presents the inference methods, the fourth presents different learning approaches and the fifth presents the conflict analysis. The sixth chapter explores some commands of R language related to bayesian concepts and inference and on seventh are presented some final considerations.
Description
Keywords
Redes bayesianas Teorema de Bayes Grafos acíclicos direcionados Inferência probabilística Bayesian networks Bayes theorem Directed acyclic graphs Probabilistic inference Matemática, Estatística e Aplicações . Faculdade de Ciências Exatas e da Engenharia