CAREER: Automatic Synthesis of Bidding Strategies for Trading Agents

North Carolina State University, Raleigh NC

Investigators

Abstract

In this project, trading agents---software programs that participate in markets---will be designed to ascertain the rules of auctions of interest and dynamically construct a decision representation. A strategy generation engine is the component of a flexible trading agent that converts the inputs (user preferences, auction rules, and models of other agents) into a decisionable format. Game theory and Markov Decision Processes are techniques that will be applied to making decisions in these markets. Previous research in the area of trading agents assumes that the market configuration is predetermined. However, the Internet marketplace is far more fragmented; a particular product will often be offered for sale in a variety of auction formats. Thus, flexible trading agents are necessary. The software programs will be made publicly available as Web-based learning materials for e-commerce courses. These materials will enable instructors in e-commerce and artificial intelligence courses to use trading agent games for class projects.