Truth and Learning in Multi-Agent Markets

In this paper, we propose an algorithm for improving the ability of decision making of buying and selling agents in an agent-based electronic marketplace. In proposed model, Selling agents use k-nn learning to adjust the first bid for new buying agents based on their similarity with the past buyers. Each selling agent learns to evaluate the reputation of buying agents based on their profits for that seller and uses this reputation to dedicate discount for reputable buying agents. Also they alter their bids in order to satisfy the buying agent's preferences. In contrast, buying agents learn to model the truth of selling agents to specify that how much they can rely on selling agents’ bids. Also buying agents evaluate the reputation of selling agents based on three different factors: reputation on quality, price and delivery-time and avoid interacting with disreputable ones. The proposed model has been implemented with Aglet and tested in a large-sized marketplace. The results show that selling/buying agents that use the proposed algorithms in this paper obtain more satisfaction rather than the other selling/buying agents

کلیدواژه ها:

Truth ، K-nn Learning ، Reinforcement Learning ، Electronic Commerce Agents