Interactions-based risk clustering methodologies and algorithms for complex project management

Project risks have never been so present. First, projects are dealing with bigger stakes and facing stronger constraints. Moreover, projects must cope with an ever-growing complexity. Risks have then increased in number and criticality. Lists of identified project risks thus need to be decomposed, for smaller clusters are more manageable. Existing techniques are mainly mono-criteria, based on risks parameters such as nature or criticality value. Limits have appeared since project risk interactions are not properly considered. Project interdependent risks are indeed often managed as if they were independent. We thus propose an interactions-based clustering methodology with associated tools and algorithms. Our objective is to group risks, so that the interaction rate is maximal inside clusters and minimal outside. The final objective is to facilitate the coordination of complex projects by reducing interfaces when dealing with risks. We first model project risk interactions through binary matrix and numerical matrix representation. Then we define our objective function. A linear programming algorithm and two approximate iterative ones are then presented. Possible refinement through the concept of interactions similarity is also proposed. A case study in the entertainment industry is finally presented, providing us information and points of comparison for global conclusions and perspectives.