電子商務架構下的協商機制逐漸受到矚目，電腦化協商支援系統的需求也日益殷切。在各種協商支援項目中，對手喜好預測無疑是最重要的一種。因為如能獲知對手喜好（無論部分或全部），就可能主導全局，甚至創造雙贏。本研究以電子化協商為基礎，發展兼具效率、準確性與實用性的對手喜好預測方法。首先，我們結合加總計分模型與Faratin 的策略模型，做為將對手協商行為量化的依據，並明確定義與對手喜好預測直接相關之待解問題。接下來，我們以基因演算法為基礎，發展求解方法。由於求解的速度與準確性可藉由演化的代數來控制，使用者可在協商中依據臨場需求設定預測品質。此外，前人研究中對於議題數、協商戰略參數與效用函數等參數的諸多限制，在本研究中也可大幅放寬。模擬實驗結果顯示：當雙方各議題出價區間相同時，使用預測的一方平均可增加15.3%的獲利；若雙方均使用預測技術，則效用分別可增加14.6%與10.7%。當議題出價區間僅有80%重疊的狀況下，預測的一方也可增加14.3%的平均效用；若雙方相互預測，則效用分別會增加11.8%與6.2%。值得注意的是：雖然使用預測的一方效用增加，但對手的收益並未因此減損，顯示預測的一方有能力主導雙贏。

Recently, how to provide a proper negotiation mechanism in e-business (e-negotiation) is becoming an important issue. In view of the requirement of fullyor semi-automated e-negotiations, an efficient computer-aided negotiation support system is attracting much more attentions than before. At present, diverse negotiation supports have been proposed for assisting the negotiator in making decisions. Among these supports, predicting the opponent's preferences is undoubtedly the most important one. With the help of an accurate prediction, a participant can predominate in negotiation and even discover a win-win solution. In this paper, to provide an efficient and accurate prediction in negotiation, a new method is proposed to model the opponent's preferences. First, the additive scoring model and the Faratin's strategy model are combined to quantify a negotiator's preferences. By this arrangement, the prediction problem can be reformulated as an optimization problem and be solved by the genetic algorithm. By means of the genetic algorithm, the quality and accuracy of prediction can be controlled by the evolving generation in finding the solution. In addition, the restriction on the preference model in the previous work can be greatly relaxed. To demonstrate the effectiveness of the proposed method, elaborated-designed simulation experiments are performed. The results show that the profit of using the prediction method can be increased by 15.3%. When both sides apply prediction in negotiation simultaneously, 14.6% and 10.7% on the increment of profits can be achieved respectively. It is worth noting that, when only one side applies preference prediction, the profit of his opponent is not decreased. It shows that a win-win situation is actually able to be obtained by the means of the opponent's preference prediction.