Including structural factors into the metrics-based code smells detection

Abstract

Code smells help to discover and describe deeper problems in software design. Several automated methods of smell detection are based the analysis of a combination of code-related metrics relevant for a given flaw. However, some smells reflect more complex issues and require a holistic perspective that woudl cover a number of different sources of data. In this paper we experimentally verify the usefulness of including structural factors into a metrics-based detection of God Class and Brain Class code smells.