A COUPLING AND COHESION METRICS SUITE FOR

The increasing need for software quality measurements has led to extensive research
into software metrics and the development of software metric tools. To maintain high
quality software, developers need to strive for a low-coupled and highly cohesive
design. One of many properties considered when measuring coupling and cohesion is the
type of relationships that made up coupling and cohesion. What these specific
relationships are is widely understood and accepted by researchers and practitioners.
However, different researchers base their metrics on a different subset of these
relationships.
Studies have shown that because of the inclusion of multiple subsets of relationships
in one measure of coupling and cohesion metrics, the measures tend to correlate among
each other. Validation of these metrics against maintainability index of a Java program
suggested that there is high multicollinearity among coupling and cohesion metrics.
This research introduces an approach of implementing coupling and cohesion
metrics. Every possible relationship is considered and, for each, addressed the issue of
whether or not it has significant effect on maintainability index prediction. Validation of
orthogonality of the selected metrics is assessed by means of principal component
analysis. The investigation suggested that some of the metrics are independent set of
metrics, while some are measuring similar dimension.