Learning algorithms for single-instance electronic negotiations using the time-dependent behavioral tactic

Abstract

Negotiator often rely on learning an opponent's behavior and on then using the knowledge gained to arrive at a better deal. However, in an electronic negotiation setting in which the parties involved are often unknown to (and therefore lack information about) each other, this learning has to be accomplished with only the bid offers submitted during an ongoing negotiation. In this article, we consider such a scenario and develop learning algorithms for electronic agents that use a common negotiation tactic, namely, the time-dependent tactic (TDT), in which the values of the negotiating issues are dependent on the time elapsed in the negotiation. Learning algorithms for this tactic have not been proposed in the literature. Our approach is based on using the derivatives of the Taylor's series approximation of the TDT function in a three-phase algorithm that enumerates over a partial discretized version of the solution space. Computational results with our algorithms are encouraging.

Learning about the opponent in automated bilateral negotiation: a comprehensive survey of opponent modeling techniques

“…Minimize negotiation cost [7][8][9]11,50,72,73,90,103,113,137,143,146,149,151,153,155,160,165,166,183,184,188,[205][206][207] In general, it costs time and resources to negotiate. As a consequence, (early) agreements are often preferred over not reaching an agreement.…”

Towards a Quality Assessment Method for Learning Preference Profiles in Negotiation

“…The third component provides criteria for selecting negotiation strategies that should be used by negotiating agents in the tournament. Since learning of an opponent's preference profile in single-instance negotiations has to be accomplished with only the observations of the opponent's negotiation moves [8,12,17,22], typically such learning algorithms use assumptions about an opponent's behaviour. For instance, in [1,8,22] a concession assumption is used which states that negotiators on average decrease the utilities of offers as time passes in order to find a deal.…”

“…Work in the area of opponent modelling in negotiation has resulted in a variety of approaches that usually focus on learning one aspect of the negotiation process. The range of negotiation aspects that are learned includes reservation values [21], issue priorities (or weights associated with negotiated issues modelling the relative importance of each issue; [2,9]), and negotiation strategies [12,14].…”

Knowledge discovery for adaptive negotiation agents in e-marketplaces

“…As a summary, there are variety of approaches of negotiation knowledge discovery such as case-based reasoning [3], fuzzy rules [6], time series approximation [24], Bayesian learning [5,40], Markov Chain Process [25], evolutionary learning [19], constraint satisfaction [38], etc. Generally speaking, these learning approaches can be classified into the broad categories of parametric [3,19,24] or non-parametric methods [5,25,40].…”

“…Generally speaking, these learning approaches can be classified into the broad categories of parametric [3,19,24] or non-parametric methods [5,25,40]. The negotiation knowledge discovery method illustrated in this paper is based on non-parametric approach since heterogeneous negotiation agents utilizing various tactics may be deployed to e-Marketplaces.…”