Abstract
In P2M program management, managers are forced to control multiple projects which progress at the same time. However, it is difficult for them to manage all the projects by their hands. We tried to apply machine learning to program management so that we can get predictions of project failure. Due to the trial, we con-firm the accuracy has reached a level where it can be used in actual operations. On the other hand, it turns out that predicting failure is not enough to decide the risk will affect program’s failure and take some actions for its early recovery. This study aims at supporting program managers to deal with program risks. We systematically extract risk triggers which lead system development projects into trouble, and validate them by comparing several case studies.
