A quantitative cohesion complexity measure to enhancing software quality

摘要

This paper proposes a quantitative approach to measure module cohesion. The relatedness of elements within a module is quantified in the form of cohesion complexity. We first identify variable relatedness using variable dependence graph. Cohesion complexity is then analyzed and mathematically formulated in accordance with standard definitions. Variable relatedness being analyzed are data, selection, and loop. As such, traditional ordinal measure can be objectively clarified to distinguish the differences of design cohesion classification, reflecting the desired software quality. The result so obtained will help developers achieve better cohesive design of software.