A Web-Based Agent Service Platform for E-Commerce Applications

Internet-based auction is a profitable, exciting and dynamic part of E-commerce. However, the lack of standard on negotiation protocols between buyers and auctioneers makes the full automation of E-commerce infeasible. Hence, we (1) propose a web-based agent platform for E-commerce which allows humans and software agents to perform automatic auctions over the Internet; (2) develop a set of negotiation protocols based on advanced agent technologies; and (3) design Markov models to evaluate the set of negotiation protocols. Currently, our platform supports three types of auction including the English, Dutch, and Double auction. The platform architecture is componentbased which provides a robust, reliable, scalable and easy to manage environment. Agents act as software components and run in a flexible and dynamic environment for Internet-based auction. They employ the set of negotiation protocols provided by the platform to conduct neogotiations auctomatically with minimal human efforts. People can also use the platform to conduct agent simulations to predict the price trajectory of an item. Simulation results provide a good reference for sellers to choose the best auction closing time so as to maximize their profit. Moreover, we use Markov chains to model three types of auction including