Simulation and analysis of bid evaluation behaviors for multiattribute reverse auction

Abstract

Reverse auctions with multiattributes are widely used for centralized procurements of large enterprises and government departments. The bid evaluation mechanism is a key factor for a fair auction and resource allocation. The current bid evaluation process is usually done by two groups, a business expert group and a technical expert group, separately. Taking bounded rationality as a precondition, we analyze the existence of antagonistic feelings between the two groups. The antagonistic feelings may grow as time goes by. To verify the discovered points, a multiagent-based approach is used to simulate the bid evaluation behaviors. The prototype system of a current bid evaluation mechanism is described by a multiagent model. In the model, the behaviors of both auctioneer and bidders are described by the attributes of each agent and decision rules. Their interactions are described by the collaboration and information exchange between different agents. The procedure of multiagent simulation is specifically presented. Then, the multiagent simulation system is developed and repeatedly runs on the platform of MATLAB GUI while adjusting relevant parameters. The results demonstrate that the behaviors of raising and repressing bid scores are increasingly severe with the passage of time. The result is that differences between the two groups increase with a decrease in the objectiveness of experts. The accumulation of antagonistic feelings can be obviously observed. The current mechanism is unfair to the bidders, as it may make the best bid lose the auction. To solve this problem, two modified bid evaluation mechanisms are suggested. The simulation comparison results show that the new bid evaluation mechanisms can efficiently weaken the antagonistic feelings of the two groups. Finally, three constructive suggestions are put forward to the organizer of tendering and bidding.