The behavior analysis of grouped multi-attribute auction based on multi-agent

Abstract

Reverse auction with multi-attributes is widely used for centralized procurements of large enterprises and government, and grouped multi-attribute is the popular bidding evaluation mechanism at present. Bidding evaluation mechanism is a key factor on the fairness of bidding and resource allocation. On account of the defects including distortion about man-made evaluation, strong psychology intervention and so on, taking bounded rationality as precondition, we forecast the existence about the accumulation of antagonism between two groups of bid evaluation experts. We apply the multi-agent theory to research bidding evaluation behavior, prototype system of the current mechanism is described, meanwhile the multi-agent model is structured. Then we define the attribute of each agent, decision rules and the way in which different agents use for exchange and collaboration. The procedure of multi-agent simulation is given specifically. Then we realized the analytical system of bidding evaluation based on matlab GUI, simulated repeatedly adjusting relevant parameters. The results demonstrate that raising and pressing the grades is increasingly severe with time gone, the difference between two groups increasingly and objectiveness worse and worse, the accumulation of antagonistic feelings is obvious. It certainly will make the auction possibly lose the best bid owing to the unfair bidding evaluation mechanism. To solve this problem, we suggest three improved proposal. The next step, it is an urgent demand to study deeply and optimize the current bidding evaluation mechanism with the help of of multi-agent.