Monitoring Software Quality by Means of Simulation Methods

Daniel Honsel, Verena Herbold, Marlon Welter, Jens Grabowski, Stephan Waack

Abstract

The evolution of software projects is driven by developers who are in control of the developed artifacts and the quality of software projects depends on the work of participating developers. Thus, a simulation tool requires a suitable model of the commit behavior of different developer types. In this paper, we present an agent-based model for software processes containing the commit behavior for different developer types. The description of these types results from mining software repositories. Since relationships between software entities, e.g., files, classes, modules, are represented as dependency graphs, simulation results can be assessed automatically by Conditional Random Fields (CRFs). By adjusting simulation parameters for one project we are able to give a quality trend of other projects similar in size and duration only by changing the effort and the size of other projects to simulate.
Keywords: 
Agent-Based Simulation; Mining Software Repositories; Software Dependency Analysis; Conditional Random Fields
Document Type: 
Articles in Conference Proceedings
Booktitle: 
Proceedings of the 10th ACM/IEEE International Symposium on Empirical Software Engineering and Measurement
Series: 
ESEM '16
Address: 
New York, NY, USA
Publisher: 
ACM
Pages: 
11:1-11:6
Year: 
2016
URL: 
http://doi.acm.org/10.1145/2961111.2962617

Main menu 2

2011 © Software Engineering For Distributed Systems Group