Code Clone Detection by Evaluating Combinations of Software Metrics

Abstract

Code cloning can be defined as an act of reusing a code segment by copying it from one section of the software and then pasting it with or without some slight alterations into other sections of the software. It is a basic means of software reuse. Most of the large software systems consist of numerous code clones because code cloning has been extensively used within the software development design community. Although it is easy to develop softwares by applying code cloning but cloning is problematic for both maintenance and quality of a software system. Code cloning results in increasing maintenance effort because during the maintenance phase of a software if an error or bug is found in one code fragment, then all its corresponding clones should be find out to detect and correct the same error or bug. Because of the problems caused due to code cloning, there is a need to detect code clones and the detection process is known as clone detection. A number of code clone detection techniques have been proposed so far. In this thesis, an approach is proposed for selecting a set of relevant metrics for detecting code clones using metrics based code clone detection techniques. The proposed approach evaluates a set of independent metrics and these metrics are evaluated on the basis of their precision and recall values in clone detection starting from all combinations of one metric and then gradually increasing the number of metrics in the metrics combinations until the complete set of metrics involved in the approach are evaluated. The result of implementing the proposed approach on a C language software system is provided as example.