The Conceptual Coupling Metrics for Object-Oriented Systems (2006)

Author(s): Denys Poshyvanyk and Andrian Marcus
Venue: ICSM IEEE International Conference on Software Maintenance
Date: 2006

Type of Experiement: Case Study

Quality
3

Details
Chose ten open-source projects from sourceforge.net in order to test the new CoCC (Con-
ceptual Coupling of Classes) metrics on. Tested them in comparison with CBO, RFC, MPC,
DAC, ICP, ACAIC, OCAIC, ACMIC, and OCMIC in order to see if CoCC can capture a
new dimension in coupling measurement.

Summary
Poshyvanyk and Marcus propose a new coupling metric CoCC that will use information
retrieval in order to evaluate the similarity between two classes. Using a parser to parse
identifiers and comments this is then passed through a Corpus builder which is then inter-
preted by a latent semantic indexing tool that constructs vectors which can then be analyzed
for coupling between classes. This metric captures a whole new dimension of software metrics
outside of the actual architecture of the code. This captures the conceptual and cognitive
relationship between classes through its identifiers and comments.
They conclude with their metric being able to add to the existing mainstream metrics
as described and used in comparison within the article. They do not include inheritance or
dynamic features within their proposed coupling metric.

0