Mining quality reputation of developers from a revision control system - and what it is good for

The internal quality of source code is in uenced by various factors of the development context, e.g. programmers. We extract authorship information about collaboratively written code from a revision control system and combine it with data about le quality in order to train a reputation system. The developers' reputation is then used to predict the quality of other les they contributed to. This paper validates the existence of a relationship between developers and code quality, shows that reputation is a sound concept, compares dierent authorship mining algorithms, and measures predictive capabilities. In the future, the results may serve as benchmarks for authorship mining and help to coste ciently increase the quality of software: by enabling reputation systems that redu ce cowboy coding, or