Abstract
A lot of requirements are discarded throughout the product development process. However, resources are invested on them regardless of their fate. If it would exist a model that predicts reliably and early enough whether a requirement will be deployed or not, the overall process would be more cost-effective and the software system itself more qualitative, since effort would be channeled efficiently. In this work we try to build such a predictive model through modelling the lifecycle of each requirement based on its history, and capturing the underlying dynamics of its evolution. We employ a simple classification model, using logistic regression algorithm, with features coming from an engineering understanding of the problem and patterns observed on the data. We verify the model on more than 80,000 logs for a development process of over 10 years in an Italian Aeronautical Company. The results are encouraging, so we plan to extend our study on one side collecting more experimental data and, on the other, employing more refined modeling techniques, like those coming from data mining and fuzzy logic.