Research On Argument Mechanism Of Argumentation Based Agent Negotiation

With the continuous development of automated negotiation that based on Agenttechnology, because of the advantages of argumentation-based negotiation, itgradually becomes the focus of the automated negotiation research field. Thisresearch was conducted in the background of the argumentation-based negotiationthat use Agent technology in e-commerce environment, and choose the evaluation andgeneration of proposal and argument as the main research questions. The detailedresearch contents are as follows: In the first part, we give the description of the overallprocess of the argumentation-based negotiation that designed by this article, on thebasis of this process, we carried out a detailed research about the evaluation methodgenerarion rules of argument and proposal. In the second part, we studied theevaluation method of proposal and argument in detail. In this part, evaluation methodincludes two parts: argument evaluation and proposal evaluation, in the argumentevaluation process, we use fuzzy comprehensive evaluation method to evaluateargument and historical negotiations record, then we get the argument evaluationvalue, in the proposal evaluation process, we use multi-attribute utility theory toevaluate the proposal, then we get the proposal evaluation value, then according tothis two values, we get an comprehensive evaluation value, and then compared to thethreshold, to determine Agent’s reaction. In addition, we use Q-learning algorithm tolearn the evaluation threshold, make the evaluation threshold dynamically adjustedwith the negotiations carried out, to improve the efficiency and success rate of thenegotiations, in the third part, we study the generation process of proposal andargument. In this part, generation rules includes two parts: argument generation andproposal generation. In the argument generation process, we use similarity formula togenerate the anti-argument, in the proposal generation process, according to thecase-based reasoning and similarity formula, Agent generates a correspondingproposal, then Agent send the argument and proposal to the other Agent, at this point,we complete the generation process. In this article, we implemented the two keyaspects of the argumentation-based negotiation: evaluation of proposal and argument,generation of proposal and argument. Finally, we use.NET platform to design aprototype system for evaluation methods and generation rule. The evaluation methodand generation rule presented in this paper can complete the interaction of the argumentation-based negotiation, which prompte the negotiations to develop toward amore practical direction.