EMPIRICAL ANALYSIS OF OBJECT-ORIENTED COUPLING AND COHESION MEASURES IN DETERMINING THE QUALITY OF OPEN SOURCE SYSTEMS

摘要

Design metrics play an important role to help developersrnunderstand the software design aspects which later can bernused to improve software quality. This paper providesrnempirical evidence supporting the roles of object-orientedrn(OO) coupling and cohesion metrics, using a subset ofrnChidamber and Kemerer OO metrics suite, in predictingrnthe post-delivery defects. The findings, based on variousrnopen source systems developed in Java, indicate that, arncoupling measure, Coupling between object classesrn(CBO) is very highly significant in predicting defectsrnwhile Lack of cohesion metric (LCOM), as a cohesionrnmeasure is not significant. Even after controlling for size,rnCBO metric is still significant as a predictor for postdeliveryrndefects.