My research focused on how to make a negotiation software agent predict, learn and better respond to opponent's behaviour in negotiation. To be applicable to negotiations in real world, this learning approach tries to make less assumptions about the opponent, and makes use of only the opponent's previous offers to predict their behaviour and respond with appropriate offers, in order to get better share in the final deal or agreement.

Keys: Negotiation, Policies, Tactics and Strategies, Learning in Multi-Agent System, Game Theory, Cooperation and Competition in e-Commerce