For software project management, it is very important to identify risk factors which make project into runaway. In this study, we propose a method to extract improvement action items for a software project by applying association rule mining to the software project repository for a metric of “cost overrun”. We first mine a number of association rules affecting cost overrun. We then group compatible rules, which include several common metrics having different values, from the mined rules and extract improvement action items of project improvement. In order to evaluate the applicability of our method, we applied our method to the project data repository collected from plural companies in Japan. The result of experiment showed that project improvement actions for cost overrun were semi-automatically extracted from the mined association rules. We can confirm feasibility of our method by comparing these actions with the results in the previous research.