ABSTRACT It is quite frequent to copy and paste code fragments in software development. The copied source code is called a software clone and the activity is referred to as code cloning. The presence of code clones hamper maintenance and may lead to bug propagation. Now-a-days, model driven development has become a standard industry practice. Duplicate parts in models i.e. model clones pose similar challenges as in source code. This paper presents an approach to detect clones in Unified Modeling Language class models. The core of our technique is the construction of a labeled, ranked tree corresponding to the UML class model where attributes with their data types and methods with their signatures are represented as subtrees. By grouping and clustering of repeating subtrees, the tool is able to detect duplications in a UML class model at different levels of granularity i.e. complete class diagram, attributes with their data types and methods with their signatures across the model and cluster of such attributes/methods. We propose a new classification of model clones with the objective of detecting exact and meaningful clones. Empirical evaluation of the tool using open source reverse engineered and forward designed models show some interesting and relevant clones which provide useful insights into software modeling practice.
