Designing Of Multi-attribute Auction Model Based On MAS

With the rapid development of the Internet, online auction becomes one of the hot topics. At present, online auction only pays attention to individual attribute. However, a number of other attributes should be taken into consideration as well. Meanwhile, it is similar to the computerization of traditional auction and lack of intelligence and automation. Multi-agent techniques have been employed to solve these problems. This thesis analyses the theory and the status quo of multi-attribute auction based on agent, furthermore, it studies the multi-attribute auction model, the mechanism of coalition allocation and the strategy of utility allocation.A multi-attribute auction model based on the multi-agents is designed. This model uses agent group to replace the participants of both sides and intermediary agent group to reach the agreement. Three algorithms used in the model are proposed. The algorithm of optimal bidding in multi-attribute auction provides the prominent bid for purchasers according to the weight of attributes. Using collaborative filtering technology, the biding recommendation algorithm based on similarity of attributes can reduce the times of negotiation and increase the probability of auction. The one-to-many multi-attribute negotiation algorithm has the characteristics of parallelism and fair. It can support multi-attribute negotiation and guarantee the system to reach consistent agreement. Experiments and analysis show the efficiency of these algorithms.A coalition formation mechanism based on multi-attribute auction is proposed. The mechanism can form coalitions dynamically and rapidly and increase the value of the system. A new coalition formation strategy based on bargaining is established. It can realize the agent utility allocation according to self-expression in coalition and promote the agent to form a stable coalition again in the long run. The effect of thisapproach has been demonstrated.