An improved behavioral matching for anti-pattern based abstract factory recommendation

For a developed project, Abstract Factory can be recommended using structural and behavioral matching between the defined anti-patterns and project code. This paper proposes a refinement of the existing behavioral matching technique, where all possible aspects of Abstract Factory behaviors are considered. This matching phase is enhanced by generating family matrix using conditional and action parse, and product type matrix using class association. Those two matrices are analysed to identify whether a pair of classes have the same product type. If no such matching is found, regeneration of family matrix by class wise parsing is done and again the matrices are analysed for recommending Abstract Factory. The approach is implemented in the form of a tool named Source code based Abstract Factory Pattern Recommender (SAFPR). The performance is justified by an experimental result analysis that showed, SAFPR is successful in Abstract Factory recommendation for all the dataset projects.