Decision support for preference elicitation in multi-attribute electronic procurement auctions through an agent-based intermediary

Abstract

Multi-attribute auctions have become increasingly popular in enterprise procurement. In the auctions, the elicitation of the preference of an auctioneer concerning multiple attributes is a central task in determining the winner(s). Considering the difficulty of explicit elicitation, a preference elicitation framework is proposed to assist the auctioneer in inferring his/her underlying preference model(s). The auctioneer is expected to provide the information of attribute weights, the holistic preference relations concerning a set of reference bids and the comparison information of intensities of preferences between some pairs of bids on all attributes and/or a particular attribute. Based on this information, a linear programming model is constructed to infer the preference model(s) of the auctioneer so that the estimations are as consistent as possible with the given preference statements. Furthermore, a method is also given to select a representative preference model from the set of compatible ones. The framework is implemented by an intelligent buyer agent called e-buyer which has five main components, i.e., a semantic analyzer, a preference elicitation module, a bid evaluation module, a model base, and a database. The e-buyer is embedded into an auction intermediary, and the proposed preference elicitation models are stored in the model base. Several graphical user interfaces are also presented to visualize the future trading intermediaries. Finally, a numerical example is given to illustrate the framework and show the effectiveness of the preference elicitation models.