為處理單一決策者及群體決策環境下決策者風險偏好的影響，本研究運用增量分析的概念一般化TODIM法(葡萄牙語 Interactive and Multi-criteria Decision Making 的意思)，並擴展至群體決策的環境中，以建構一整合式的多準則群體決策支援模式。本研究除修改傳統TODIM價值函數在損失部分的缺陷外，亦對不同類型準則分別使用線性與非線性價值函數，也針對運用增量分析法後決策者對方案之排序以及群體中每位決策者的相對權重進行整合。其次，對不同序數值轉換為基數值的方法與決策者排序距離矩陣之特徵向量以決定決策者相對權重的方法進行比較與討論。另外，經由模式內價值函數之參數敏感度分析以及不同決策者權重比較也證明了本模式的穩固性。最後，案例說明顯示本整合架構之可行性。 This study aims to generalize TODIM (an acronym in Portuguese of Interactive and Multi-Criteria Decision Making) by incremental analysis (IA) for a risky decision making and extending it to a group decision-making environment. For generalization, two types of scaling effects are overcome. One effect is due to the defect of original TODIM formulation in losses part of the two parts value function, and the criteria with smaller weights will contribute larger dominance values in the computing process. The other effect is derived from aggregating partial dominance measurements among different characteristics of criteria in TODIM while the criteria can be divided into two categories, benefits and costs, for effective resource allocation. In such a way, different types of value functions are considered for reflecting the decision maker’s risk preference. IA is then employed to rank alternatives according to the given cutoff benefit-cost ratio for two accumulated dominance measurements. A fuel buses example is illustrated. The proposed method is extended to a group decision making environment in which multiple decision makers (DMs) execute the model. The ordinal rank of each DM can be converted to cardinal value through rank sum and regression-like function. In addition, the relative decision power of each DM is taken into account based on the eigenvector of their ranking distance comparison matrix. Later, sensitivity analyses are employed on the different values of the parameters in the value function, the cutoff benefit-cost ratios, and different weights of DMs to demonstrate the robustness of the proposed model for group decision. Furthermore, a number of other MCDM techniques have been compared. The results show that the proposed model is feasible and effective for the demonstrated example under risk.