Bellwethers: A Baseline Method For Transfer Learning

PDF

Software analytics builds quality prediction models for software projects. Experience shows that (a) the more projects studied, the more varied are the conclusions; and (b) project managers lose faith in the results of software analytics if those results keep changing. To reduce this conclusion instability, we propose the use of "bellwethers": given N projects from a community the bellwether is the project whose data yields the best predictions on all others. The bellwethers offer a way to mitigate conclusion instability because conclusions about a community are stable as long as this bellwether continues as the best oracle. Bellwethers are also simple to discover (just wrap a for-loop around standard data miners). When compared to other transfer learning methods (TCA+, transfer Naive Bayes, value cognitive boosting), using just the bellwether data to construct a simple transfer learner yields comparable predictions. Further, bellwethers appear in many SE tasks such as defect prediction, effort estimation, and bad smell detection. We hence recommend using bellwethers as a baseline method for transfer learning against which future work should be compared
Submitted 17 Mar 2017 to Software Engineering [cs.SE]
Published 21 Mar 2017
Updated 21 Jan 2018
Author comments: 23 Pages
http://arxiv.org/abs/1703.06218
http://arxiv.org/pdf/1703.06218.pdf

0 comments