Mining Software Dependency Networks for Agent-Based Simulation of Software Evolution
Authors:
- Verena HonselVHVerena Honsel
- Daniel HonselDHDaniel Honsel
- Steffen HerboldSHSteffen Herbold
- Jens GrabowskiJGJens Grabowski
- Stephan WaackSWStephan Waack
- 1 November 2015
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 102-108
Abstract
During the software development process, the time and resources for quality assurance are limited. Therefore, project managers benefit from knowing in advance if a decision leads to decreasing quality. For this, we build an agent-based simulation tool for software processes for testing the effect of changing parameters, e.g., development team size. Since often changed software entities tend to be more defect-prone, we analyze the evolution of common file changes and evaluate its applicability for our agent-based simulation. For the estimation of simulation parameters we performed a case study focusing on change coupling dependency graphs of open source software projects. The analysis of this also provided valuable insights in the structure of these dependencies. By comparing empirical observations with simulation results we support the assumption that file dependencies can be simulated. Moreover, we are able to reproduce the observed patterns with a parameter set from another project which, therefore, indicates the transferability of the simulation for projects similar in size and duration.Keywords
This publication has 19 references indexed in Scilit:
- Complex adaptive systems modeling with Repast SimphonyComplex Adaptive Systems Modeling, 2013
- Node centrality in weighted networks: Generalizing degree and shortest pathsSocial Networks, 2010
- Community detection in graphsPhysics Reports, 2010
- Predicting defects using network analysis on dependency graphs2008
- The influence of organizational structure on software quality2008
- Analysing Software Repositories to Understand Software Evolution2008
- Agent-based simulation of open source evolutionSoftware Process: Improvement and Practice, 2006
- Predicting risk of software changesBell Labs Technical Journal, 2002
- Visualizing software changesIEEE Transactions on Software Engineering, 2002
- The Kolmogorov-Smirnov Test for Goodness of FitJournal of the American Statistical Association, 1951