協商中的對手喜好預測與協商策略應用

Opponent's preferences prediction and strategies develop in E-negotiation

葉學杰(Hsueh-Chieh Yeh)

指導教授 ： 張昭憲

淡江大學/管理學院/資訊管理學系碩士班/碩士(2012年)

https://doi.org/10.6846/TKU.2012.01047

摘要

電子化協商的發展方興未艾，各種協商支援系統也紛紛被提出。在眾多的支援項目中，最受人矚目的莫過於對手喜好預測。若能了解對手喜好，便有機會主導協商流程，達成預設目標。學者們雖然提出了許多對手喜好預測方法，但多以精準預測為目標，鮮少考慮當了解對手喜好之後，己方該以何種策略來因應，使預測的積極效果大打折扣。有鑑於此，本研究致力發展一套具有動態策略調整能力的對手喜好預測方法，其能根據預測結果，在協商中導引雙方的出價過程，取得預設的協商結果。為達成此目的，我們首先以Faratin提出的協商模型為基礎，配合基因演算法，預測對手可能的喜好。為了描述協商者的態度，我們設計了多種目標函數，以引導協商流程的進行。當進行動態策略調整時，我們則以預測結果為基礎，透過內部模擬協商，找出可行的策略調整方式，以符合目標式的規範。為驗證方法的有效性，本研究進行了大量的模擬實驗。實驗結果顯示:有預測的一方，無論協商者選擇何種策略目標，均能提升己方的效用值。當雙方均使用喜好預測，且均以自利（selfish）為目標時，可得到最佳的效用總合。若選擇互利或慷慨為遠程目標，則可降低協商回合數，並提高協商成功率。當配合動態策略調整，則效果可獲得進一步提升，顯示動態策略調整的必要性與有效性。

關鍵字

對手喜好預測 ； 協商支援 ； 協商戰略與策略 ； 自動化協商

並列摘要

The development of E-Negotiation is still growing as various Negotiation Supporting Systems are also being offered one after another. Among various Supporting Systems, nothing is getting more attention than opponent’s preferences predictions. If we understand opponent’s preferences, we can get the chance to lead the negotiation process and achieve the preset goal. Though scholars have offered many methods to predict opponent’s preferences, most of them focus on accuracy, rather than knowing what tactics to respond one understood their preferences which end up reducing the effects of predictions. In regard to this fact, this research dedicates to develop opponent’s preference prediction which have the ability of dynamic tactics, it can attain the predicted negotiation outcome. In order to attain this goal. First, we use the negotiation model which put forward by Faratin as base, working in coordination which Genetic algorithm to predict possible opponent’s preference. To describe negotiation’s attitude, we design several object function to guide the process of negotiation. When doing the adjustment of dynamic tactics, we find out practicable method of tactic adjustment by simulative negotiation inside on the basis of predictive result to confirm with objective standard. To testing the validity of the method, this research has proceeded abundant of simulated experiments. The result of the research shows that no matter what kind of tactics’ goal the negotiation have chosen , they can an upgrade their own utility value in the side which has doing prediction. When both sides use opponent’s preferences prediction and also make selfish as their goal, it can get the best combination of utility. If we choose mutually beneficial or generosity as the long-term object, we can reduce the rounds of negotiation and raise the rate of success at the same time. The effect can obtain further promotion when matching up dynamic tactics. Then, it reveals the necessity and validity in adjustment of dynamic tactics.

並列關鍵字

Opponent's Preferences Prediction ； Negotiation support ； Negotiation Strategies and Tactics ； Automated negotiation

參考文獻

[1] 黃淳韋. 2011. "電子化協商架構下的對手喜好預測與策略運用方法," 淡江大學資管系碩士論文).

[4] Barbuceanu, M., and Lo, W.K. 2001. "Multi-Attribute Utility Theoretic Negotiation for Electronic Commerce," Agent-Mediated Electronic Commerce III), pp 15-20.

[5] Blecherman, B. 1999. "Adopting Automated Negotiation," Technology in Society (21:2), pp 167-174.

[6] Carbonneau, R., Kersten, G.E., and Vahidov, R. 2008. "Predicting Opponent’s Moves in Electronic Negotiations Using Neural Networks," Expert Systems with Applications (34:2), pp 1266-1273.

[7] Faratin, P., Sierra, C., and Jennings, N.R. 1998. "Negotiation Decision Functions for Autonomous Agents," Robotics and Autonomous Systems (24:3), pp 159-182.