Multi-attribute Automatic Negotiation Research Based On Agent

The economic development and emergence of new internet technology promotes the progress of e-commerce services. The emergence and advancement of e-commerce brings great changes and far-reaching impact on the each aspect of social life. Under the e-commerce activities, in generally, negotiation will be used to obtain the consensus of both sides to solve the bilateral dispute on trade or service. Traditional negotiation usually is face to face for the buyers and sellers, it not only costs a lot of manpower and financial resources, but also is impacted by the social or other factors and apparents some randomness. This way has not well adapted to the needs of development of the e-commerce market; particularly it can not achieve automatic processing for multi-attribute negotiation process. Agent technology can partially or completely replace human beings to resolve negotiation problems in trade, which is not restricted by the time and place, can improve the negotiation efficiency and promote e-commerce development.Above the analysis, the thesis analyzed the automatic negotiation mechanism by drawing on previous research results, designed specific protocol by the requirements of business negotiation and the characteristics of multi-attribute negotiation, which will help both sides to obtain maximum benefits. At the same time, the thesis introduced two kinds methods of solving the optimal strategy and interval number is proposed to express decision-makers'preferences. An interval-valued intuitionistic fuzzy number decision-making criterion is used to determine the preference degreement for the negotiating parties, which ensured objectivity of multi-attribute decision making. The genetic algorithm is used to search the optimized value, niche technology is proposed to improve optimization speed and prove the effect of result. Finally, taking trading data cable as an example, a simulation experiment is carried on the described problem with the different negotiation strategies--no having learning mechanism and the mechanism of with learning. Negotiation times and transaction values are counted by experiment and applicable conditions and outcome of negotiations are analyzed comparatively, so the research is proved be feasibility.