An Empirical Study of a Hybrid Code Clone Detection Approach on Java Byte Code

Code clones increase the complexity of the system; therefore the software maintenance costs. Code clone detection techniques have been proposed and evaluated based on metric value and runtime evaluations. But in the existing methods, many false positive clones are detected. In this paper, we suggest a hybrid approach combining Program Dependence Graph-based technique with Metric-based technique to improve the precision of clone detection. We conduct a case study on two open source code Java projects such as Eclipse-ant and Eclipse-JDT core to show the effectiveness of our tool. The application of this hybrid technique is then compared with the existing clone detection technique, CloneDR. The result shows that our tool increases the performance in precision, recall, false positive and false negative compared to CloneDR.