A Structural Analysis Method of OSS Development Community Evolution Based on A Semantic Graph Model

Abstract

Network structures of OSS (Open Source Software) development communities are becoming more and more complicated. Various mining techniques have been applied to the repositories of OSS communities. However, structure analysis of OSS development community evolution has not been established. In this article, we propose SCGM (Software Community Graph Model), a new class of graph models to define the OSS development community. Based on the SCGM, we propose a structural analysis method of OSS development community evolution. To automate the analysis method, a prototype system is implemented with the graph DB Neo4j. We applied the proposed method and prototype system to four major machine learning OSS communities, Caffe, Chainer, Jubatus, and Tensorflow, for over five years on GitHub. From the analysis, we discovered three novel characteristics of community evolution, 1) three layered community evolution models consisting of the Core, Semi-core and Non-core members, 2) three developer growth patterns in terms of contribution behavior, and 3) evolutional changes according to the interaction among developers, which is a major contribution of this work. Based on the experiments, we demonstrate the validity and effectiveness of the proposed method.