Predicting maintainability with object-oriented metrics -an empirical comparison

Abstract

A large number of metrics have been proposed formeasuring properties of object-oriented software such assize, inheritance, cohesion and coupling. We have beeninvestigating which of these object-oriented metrics canbe used as significant predictors for the maintainability ofsoftware. For this purpose, we have designed andconducted an empirical study based on historical datacollected from the maintenance history of a medium-sizedobject-oriented system. Unlike most related studies,indirect coupling has also been taken into account in ourwork in order to evaluate its impact. Our study uses themaintenance history of two software systems as evidencebase for linking software quality attributes to metricssuggested for object-oriented software. Our resultsindicate that size and import direct coupling metrics aresignificant predictors for measuring maintainability ofclasses while inheritance, cohesion, and indirect/exportcoupling measures are not.