Stepping towards dynamic measurement for object oriented software

Abstract

Software metrics are very helpful in measuring the different aspects of software like cohesion, coupling, polymorphism, inheritance etc. The objective of measuring software metrics are quality assurance, defect prediction, maintainability prediction, cost estimation, debugging, etc. Many authors proposed the use of static metrics for the software maintainability prediction (SMP) and were successful, but static metrics don't take into account the run-time behavior of software. Hence, to capture this dynamic behavior, dynamic metrics are necessary to be evaluated. This paper presents the empirical investigation of dynamic metrics for SMP and also compares them with static metrics. Six machine learning algorithms are used to build the prediction models for both the static and dynamic metrics. The performance of all models is compared using prevalent accuracy measures. Results show that dynamic metrics perform better than static metrics, and can be used as a sound alternative for SMP.