To Automatically Map Source Code Entities to Architectural Modules with\n Naive Bayes

Background: The process of mapping a source code entity onto an architectural\nmodule is to a large degree a manual task. Automating this process could\nincrease the use of static architecture conformance checking methods, such as\nreflexion modeling, in industry. Current techniques rely on user\nparameterization and a highly cohesive design. A machine learning approach\nwould potentially require fewer parameters and better use of the available\ninformation to aid in automatic mapping. Aim: We investigate how a classifier\ncan be trained to map from source code to architecture modules automatically.\nThis classifier is trained with semantic and syntactic dependency information\nextracted from the source code and from architecture descriptions. The\nclassifier is implemented using multinomial naive Bayes and evaluated. Method:\nWe perform experiments and compare the classifier with three state-of-the-art\nmapping functions in eight open-source Java systems with known\nground-truth-mappings. Results: We find that the classifier outperforms the\nstate-of-the-art in all cases and that it provides a useful baseline for\nfurther research in the area of semi-automatic incremental clustering.\nConclusions: We conclude that machine learning is a useful approach that\nperforms better and with less need for parameterization compared to other\napproaches. Future work includes investigating problematic mappings and a more\ndiverse set of subject systems.\n