Vers une approche d'analyse de l'impact du changement dans un système à objets

In this paper we propose an approach, both analytical and experimental, to analyze and predict change impact in object-oriented systems. We use a meta-model (PTIDEJ) to calculate the change impact. Data obtained from real systems are exploited to empirically study causality hypotheses between some software internal attributes and change impact. To evaluate our approach, an empirical study was conducted on a real system (BOAP). This study targeted a correlation hypothesis between coupling and change impact for a specific change starting from coupling metrics. The hypothesis was studied using machine-learning techniques. Results showed that import coupling is by far the most influent factor for change impact.