Application of Execution Pattern Mining and Concept Lattice Analysis on Software Structure Evaluation.
Kamran Sartipi, Hossein Safyallah · 2006
Software maintenance activities for producing a feature-rich system tend to impair the software’s structure into an unshaped and cost-prone legacy system. Thus, it is desir-able to keep track and measure the impacts of the newly added features on the structure of the software system. The proposed technique in this paper is based on extracting fre-quent patterns in the execution traces of a software system using a pattern discovery technique. The patterns represent functionalities that correspond to the feature specific sce-narios. In a further step, the generated execution patterns are distributed on a concept lattice to separate feature spe-cific patterns from commonly used patterns. The proposed technique allows for assigning software features onto the software system modules and provides a means for assess-ing the degree of functionality scattering among the system modules. Consequently, we measure the impact of individ-ual features on the structure of the system. A case study on the Unix Xfig drawing tool is used to present the accuracy of the approach.