Multicriteria analysis of factor selection in an uncertain system

Abstract

Decision makers are often requested to make a decision in which the factors of a system are not known or the evidence is not sufficient. The uncertainties resulting from imprecise knowledge may induce large decision errors. In order to make correct decisions, we should investigate sufficient information by reducing the uncertainty of the system. In this paper, we develop two functions to measure the uncertainty of nonspecificity and dependency of a system from the viewpoints of Evidence Theory and Statistics. Then, a. multi-objective mathematical programming with interval-valued cost is proposed to investigate the sufficient and necessary factors in order to obtain complete information of the concerned system. Theoretical supports are proposed with computational analysis. An illustrative example of a nuclear power plant accident is provided with satisfactory results.

摘要

決策者常常必需對一個影響因素未知或相關資料不足的系統作決策，這種資訊的不確定將導致決策上大量的誤差，爲了幫助決策者作正確的決策，我們應該藉由減少系統的不確定性來提供充分完整的資訊；因此，本研究從證據理論與統計理論的觀點提出兩個函數來測量因素在系統中的不確定性及相依性，然後建立一個成本係數爲區間值的多目標數學规劃模式，來求取影響系統之完整因素，並提供充分之系統资訊。此外，本文亦對系統複雜度做理論上之分析；最後以核能電廠意外事故爲例，結采顯示本揆式可以獲致滿意的結果。

Key Words:

• evidence theory
• correlation
• complete information
• multiobjective analysis

閫砧詞:

• 證據理論
• 相關性
• 完整資訊
• 多目標分析