A semantic entropy metric

Abstract This paper presents a new semantically‐based metric for object‐oriented systems, called the Semantic Class Definition Entropy (SCDE) metric, which examines the implementation domain content of a class to measure class complexity. The domain content is determined using a knowledge‐based program understanding system. The metric's examination of the domain content of a class provides a more direct mapping between the metric and common human complexity analysis than is possible with traditional complexity measures based on syntactic aspects (software aspects related to the format of the code). Additionally, this metric represents a true design metric that can measure complexity early in the life cycles of software maintenance and software development. The SCDE metric is correlated with analyses from a human expert team, and is also compared to syntactic complexity measures. Copyright © 2002 John Wiley & Sons, Ltd.