Designing a Multi-Agent Portfolio Management System

Abstract

The voluminous and readily available information on the Internet has given rise to exploration of Intelligent Agent technology for accessing, filtering, evaluating and integrating information. In contrast to most current research that has investigated single-agent approaches, we are developing a collection of multiple agents that team up on demand, depending on the user, the task and the situation, to access, filter and integrate information in support of user tasks. We are investigating techniques for developing distributed adaptive collections of information agents that coordinate to retrieve, filter and fuse information relevant to the user, task and situation, as well as anticipate user’s information needs. Our approach is based on (1) case-based task and situation models, (2)flexible organizational structuring, and (3) reusable agent architecture. We are currently implementing the system in the domain of financial portfolio management. While this paper focuses on the big picture, the final section will describe the current implementation and point to our work on detailed technical solutions.