Abstract:
This Ph.D. thesis addresses the analysis and improvement of
methods for extracting, filtering, and integrating versioning
system information in software engineering tasks.
The contributions include a new filtering metric called connection
strength, a tool for extracting and filtering logical dependencies,
and their integration into key class detection and software
clustering for architecture reconstruction. Experiments show that
information extracted from the versioning system improves the
identification of central classes, enhances software clustering,
and provides a reliable, language-independent source of
information for analyzing and maintaining legacy software.
