Not logged in.
Quick Search - Contribution
Contribution Details
| Type | Conference or Workshop Paper |
| Scope | Discipline-based scholarship |
| Published in Proceedings | Yes |
| Title | Replicating Mining Studies with SOFAS |
| Organization Unit | |
| Authors |
|
| Presentation Type | paper |
| Item Subtype | Original Work |
| Refereed | Yes |
| Status | Published in final form |
| Language |
|
| Event Title | 10th Working Conference on Mining Software Repositories |
| Event Type | conference |
| Event Location | San Francisco |
| Event Start Date | May 18 - 2013 |
| Event End Date | May 19 - 2013 |
| Place of Publication | Washington, DC |
| Publisher | IEEE Computer Society |
| Abstract Text | The replication of studies in mining software repositories (MSR) is essential to compare different mining techniques or assess their findings across many projects. However, it has been shown that very few of these studies can be easily replicated. Their replication is just as fundamental as the studies themselves and is one of the main threats to validity that they suffer from. In this paper, we show how we can alleviate this problem with our SOFAS framework. SOFAS is a platform that enables a systematic and repeatable analysis of software projects by providing extensible and composable analysis workflows. These workflows can be applied on a multitude of software projects, facilitating the replication and scaling of mining studies. In this paper, we show how and to which degree replication can be achieved. We investigated the mining studies of MSR from 2004 to 2011 and found that from 88 studies published in the MSR proceedings so far, we can fully replicate 25 empirical studies. Additionally, we can replicate 27 additional studies to a large extent. These studies account for 30% and 32%, respectively, of the mining studies published. To support our claim we describe in detail one large study that we replicated and discuss how replication with SOFAS works for the other studies investigated. To discuss the potential of our platform we also characterise how studies can be easily enriched to deliver even more comprehensive answers by extending the analysis workflows provided by the platform. |
| PDF File |
Download
|
| Export |
BibTeX
EP3 XML (ZORA)
|