Effective acceptance strategy using cluster-based opponent modeling in multilateral negotiation
Determining required conditions for accepting a bid has an important role in efficient negotiations. Conceding (aggressive) acceptance strategy of the agent against a conceding (an aggressive) opponent can lead to early and high utility agreement (losing the utility) especially in domains with discount factor. The literature is more focused on bilateral than multilateral negotiations, where they assume cost or utility function of the opponents or the historical data among the negotiation sessions are known to the agents. In this study, an effective acceptance strategy is proposed which only assumes the opponents’ rationality. The strategy extends the ParsAgent last moment concession strategy by developing an opponent model using k-means. The performance of the strategy is investigated against different combination of opponents consisting state-of-the-art strategies from Automated Negotiating Agents Competition (ANAC 2015 and 2016) in trilateral tournaments in different domains. The evaluation results indicate the superiority of proposed agent against the opponents in all domains, in average.
|Studies in computational intelligence|
|International Workshop on Agent-Based Complex Automated Negotiation, ACAN 2018, 13-19 July Stockholm, Sweden|
|Organisation||Intelligent and autonomous systems|
Khosravimehr, Z, & Nassiri Mofakham, F. (2020). Effective acceptance strategy using cluster-based opponent modeling in multilateral negotiation. Studies in computational intelligence. doi:10.1007/978-981-15-5869-6_6