A Coupling-Complexity Metric Suite for Predicting Software Quality

Coupling Between Objects and Cyclomatic Complexity have long been used to measure software quality and predict maintainability and reliability of software systems prior to release. In particular, Coupling Between Objects has been shown to correlate with fault-proneness and maintainability of a system at the class level. We propose a new set of metrics based on a fusion of Coupling Between Objects and Cyclomatic Complexity that can be superior to Coupling Between Objects alone at predicting class quality. The new metrics use Cyclomatic Complexity to 1) augment Coupling Between Objects counting to assign a strength of a coupling between two classes and 2) determine the complexity of a method invocation chain through the transitive relation of invocations involved in a coupling. This results in a measure that identifies objects that are coupled to highly complex methods or method invocation chains. The metrics were implemented as an Eclipse Plug-in and an analysis of two industry Java projects, ConnectorJ and Hibernate, demonstrates the correlation between the new metrics and post-release defects identified in system change logs.

Name thesis:degree_name

MS in Computer Science

Discipline thesis:degree_discipline

Computer Science

Year dc:date.available

2008

Author dc:creator

• Gray, Christopher L.

Contributors dc:contributor

• David Janzen

Identifier

10.15368/theses.2008.12

OAI identifier oai:identifier

oai:digitalcommons.calpoly.edu:theses-1001