Empirical evidence on OCL formal specification-based metrics as a predictor of fault-proneness

Abstract

OCL formal specification has been widely used for precise software modeling. In general, it is used to express constraints on design documents. As a novel approach, its usage can be extended to support effective testing, such as testing fault-prone components to improve software quality. In this paper, CK (Chidamber and Kemerer) metrics that can be extracted from OCL expressions are validated against module complexity. Moreover, our study proposes a new metric suite that can be directly extracted from OCL expressions to quantify module complexity. A weight has been assigned to each metric to reflect its importance in fault-prone component identification. Our study shows that metrics collected from OCL specifications are good predictors of the fault-prone components of a system during design time. An early knowledge of fault-prone components of the system can be useful to distribute efforts required for software development and testing activities.