Evaluating Usage Expertise Mined from Version Archives
Date
2012-10-04
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
One approach for modelling coding expertise is to quantify the knowledge accrued from the use of library functionality. This concept is known as Usage Expertise (Schuler and Zimmermann 2008).
This thesis makes three contributions. The first is a formal specification of a system which mines Usage Expertise from a version control repository in order to recommend developers for a change task. The second contribution is a comparison of the accuracy of the system measured against the oft-used Line 10 model of developer expertise. This evaluation finds that the usage model yields simultaneous gains in the accuracy and the diversity of recommendations. The third and final contribution is a qualitative study that explores the trust and behavioural tendencies of 9 software developers who were given the model reified as a software tool. The study finds Usage Expertise to be a trustworthy identifier of expertise. However, the study also finds a series of social and organizational factors that limit the efficacy of the model in real world contexts.
Description
Keywords
Computer Science
Citation
Ma, D. (2012). Evaluating Usage Expertise Mined from Version Archives (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/25979