This project investigates multiple strategies for automatic discovery of scheduling knowledge from projects. In managing large projects with resource and budget constraints, project managers are often overwhelmed with complex project conditions and the large number of available tools. When faced with several alternatives that initially appear equally desirable, making a good judgment can be tedious and often a painful process. The goal of this project is to develop a hybrid system which can induce the relationship between project characteristics and heuristic scheduling performance, and extract meaningful rules to guide managers in mining large project data. Preliminary experiments are conducted and demonstrate promising results. The modular structure of the system will facilitate the dynamic adaptation and learning to accommodate different project conditions. The resulting work will serve as a prototype system for further exploration of intelligent multi-agent architecture.