Formulation and analysis of a probabilistic uncertainty evaluation technique
| Autor: | Shrestha, Govinda B. |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 1990 |
| Předmět: | |
| Druh dokumentu: | Dissertation |
| Popis: | Reasoning with uncertainty in real complex problems is expected to be increasingly difficult as the need for participation by the experts increases. The Eigen Vector Method (EVM) is a simple technique to quantify personal assessment of uncertain situations. It, however, provides only point estimates and lacks a direct measure of accuracy. The proposed method of Confidence Interval Based Uncertainty Evaluation Technique (CIBUET) explores a method to estimate the variability along with point estimates. A brief review of uncertainty evaluation techniques, the basics of decision theory as a framework for analyzing preferences, and the method of EVM to estimate the priorities are presented. The variance of these point estimates result from the inconsistency in the judgement ratios, which is represented by introducing error factors with each judgement ratio. The nature of the error factors are discussed, and then modelled as lognormally distributed. Methods to estimate the parameters of the error factors and the variances of the priorities are developed. These results are corroborated by numerical simulation. The framework for several applications of the technique is presented. It is illustrated by analyzing a number of example problems taken from the literature. The methodology to extend the technique to analyze inconsistent data encountered in the process of power planning is developed in detail. The application is fully illustrated using the context and data from an actual power planning study. Several approaches taken to extend the technique to multi-level hierarchies are presented. Numerical simulation is utilized to investigate the validity of the approaches and select the most appropriate model. Some special features of the technique and the underlying assumptions with the subsequent implications are discussed at the end. Directions for further research to enhance and extend the technique are identified and some possible approaches for the same are outlined. Ph. D. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|