Topic Matching Based Change Impact Analysis from Feature on User Interface of Mobile Apps

The complexity of mobile applications often lies in the user interface (UI). To update function provided by UI or just fix bugs related to UI, software maintainers primarily need to obtain the location of source code implementation and detect change set. Since UI related feature is tightly related to the class containing the declaration of the UI component, this paper proposes a topic matching based change impact analysis method from feature on user interface of mobile apps. Our approach combines LDA model with program dependency to realize the change impact analysis. Considering app′s small scale and few comments, a novel preprocessing method combining tf-idf with term weight based on structural information is applied to LDA model. Experiments on 16 update records of 4 open source apps show the effectiveness of our proposed method.