A fuzzy-genetic decision support system for project team formation

The problem of optimal team formation is domestic to many areas of work organization including education, sport, and business. It is beyond manual implementation to build near optimal teams as soon as the pool of available personnel grows into several tens. The selection process itself is usually well defined — for each team we construct the criteria relating to the required properties (i.e., capabilities) of the team members. Because these properties can be arbitrarily combined in the personnel, the objective function becomes self-conflicting. This aggravates the team formation and calls for a specialized software support.In the paper we present a new fuzzy-genetic analytical model for the problem of project team formation. It builds on previous quantitive approaches, but adds several modeling enhancements like derivation of personnel attributes from dynamic quantitive data, complex attribute modeling, and handling of necessary overcompetency. We improve the flexibility of requirements specification using a special format that expresses the required team capabilities using fuzzy descriptors. We then define a single compound objective function, which incorporates multiple opposing criteria that the solution should maximize. To optimize the selection of multiple project teams with possibly conflicting requirements, we propose a special adaptation of island genetic algorithm with mixed crossover where the fitness of common solution is used to drive the selection within the islands. We test the effectiveness of the system using artificial domains of different complexity and describe some practical experiences of using the system in the educational process.