Combining Information Retrieval Methods to Improve Automated Traceability Recovery

The intensive human effort needed to manually manage traceability information has increased the interest in utilizing automated traceability recovery techniques. Promising results have been achieved using Information Retrieval (IR) techniques to identify candidate traceability links between different types of artifacts—e.g., requirements, use cases, code classes. Different IR methods have been proposed to support such a task but until now there is no single method that sensibly outperforms the others. In this paper we propose to combine different IR techniques aiming at improving the accuracy of stand-alone IR-based traceability recovery methods. The results of a case study conducted on two software repositories demonstrate that the combination of one of the IR methods experimented with the others provides statistically significant improvements as compared with stand-alone methods. The nature of the two repositories used in the case study also allows us to investigate the influence on the recovery accuracy of the language and the type of artifacts to be traced.