Multi-agent negotiation system for class scheduling

Abstract

The current class scheduling has difficulties in reflecting students`` preferences for the classes that they want to take and forecasting the demands of classes. Also, it is usually a repetitive and tedious work to allocate classes to limited time and resources. Although many research studies in task allocation and meeting scheduling intend to solve similar problems, they have limitations to be directly applied to the class scheduling problem. In this thesis, a class scheduling system using multi agents-based negotiation is suggested. This system consists of student agents, professor agents and negotiation agents. Each agent acts in accordance with its respective human user``s preference and performs the repetitive and tedious process instead of the user. The suggested system utilizes negotiation cost concept to derive coalition in the agent``s negotiation. The negotiation cost is derived from users`` bidding prices on classes, where each bidding price represents a user``s preference on a selected class. Two experiments were performed to verify the negotiation model in the scheduling system. The result of the experiment showed that it could produce a feasible scheduling solution minimizing the negotiation cost and reflecting the users`` preferences. The performance of the experiments was evaluated by a class success ratio.