不確定數據下多屬性效用評估模式之構建──以台南捷運路網方案為例

Multi-attribute Utility Evaluation Model with Uncertain Data-The Rapid Rail Network in Tainan

多屬性效用評估法常應用在都市交通方案之評估，而多屬性效用方案評估模式構建時，對於屬性資料值常假設為一確定之數值。但未來都市發展屬性資料多為預測數據，因預測值會有估計誤差，當規劃者將未來可能發生的情境以區間範圍的屬性資訊提供給決策者參考，此時決策者對於區間數值的偏好感受會受到屬性值的變異程度及可能發生機率之影響。本研究以台南捷運路網方案之評估為例，嘗試以中間值線性方式、指定變異變數方式、平均數及標準差指數方式、考慮機率分配方式等不同的效用函數型態來處理不確定資料下之方案評估問題。當以全部樣本數建立模式時，結果顯示不確定數據之非線性效用函數型態與中間值線性函數並沒有顯著差異。若依資料變異特性將樣本數分成不同決策群組，將決策群組在參數校估過程，由模式解釋程度自動區隔出風險樂觀者、風險中立者、風險悲觀者三類決策群組，實證後發現分群後之多屬性效用評估模式的解釋能力非常好，適用在都市交通方面之評估問題。

The multi-attribute utility function frequently assumes attribute value as certainty data. Notably, urban attribute data that is estimated will have large bias. When one considers a decision using uncertainty attribute data, variability and possibility must be integrated into the utility function. To establish multi-attribute utility models with uncertainty data, this study develops several possibility explanation models with stated preference data for a transportation project. When constructing the models based on observations, the nonlinear utility function is not better than the linear function. If observations are divided into risk proneness, risk neutrality and risk aversion decision clusters by the likelihood ratio from the utility function, empirical results show the model explanatory power is good.