層級分析法成對比較基礎標度系統之研究

Abstract

決策分析的本質與目的在幫助決策者透過有系統的思考，使其決策選擇能與其價值觀與判斷一致。層級分析法(Analytical Hierarchy Process, AHP)是由沙堤教授(T. L. Saaty)所提出的一種實用決策方法，並廣泛為研究或實務運用者所使用。為了對方案進行排序擇優，AHP採用成對比較方式進行判斷，並用口語方式表達後再轉換成比例數值尺度（1-9標度值），藉以定量表示相對重要性程度。然而，成對比較的口語判斷結果卻在轉換成1-9標度值時可能出現系統性誤差，容易造成決策誤導而不知。因此本研究利用問卷與統計分析探討AHP法中兩兩比較之口語標度值與一般人之認知標度值的差異，發現確有系統偏誤，並對3~8個準則問題平均有4.7~12.5%的機率造成最佳解的標示錯誤。

The purpose of decision analysis is to achieve consistency among decision makers' choices, values, and judgments. The Analytical Hierarchy Process (AHP) was designed by T. L. Saaty as a decision making aid. In order to prioritize alternatives, the AHP used pairwise comparisons and transferred them from oral expression to numerical ratio scale, the fundamental scale (1-9), to determine their priority of relative importance. The translation from oral to ratio scale demonstrates a systematic bias that might cause errors in final ranking without decision makers' awareness. This paper used questionnaires and statistical analysis to examine this kind of bias. The results show that this bias does exist. This bias may cause 4.7~12.5% probability of failing to identify the best alternative for problems with 3~8 attributes.