Faults prediction using object-oriented software metric

Abstract

Recently, the Open Source Software (OSS) has gained popularity and has impacted the software industry at large. Many agencies, including the Malaysian government agencies are adopting open source projects due to the merit they offer. Due to the vast usage of OSS in the government administrations and many reported problems, there is a pivotal need to study on reliability and quality of the code for those applications. Thus, the attribute of the source code of these applications should be measured. This research investigates how the object-oriented metrics by Chidamber and Kemerer (CK) are used to predict the fault-proneness in the source code of the open source applications used by the Malaysian government, namely, MyMeeting, MyTaskManager and MyBooking. In this research, in order to validate the usefulness of object-oriented metrics for fault-proneness prediction, several analyses are conducted using Statistical Package for Social Sciences (SPSS) such as Spearman correlation, multiple linear regressions and univariate logistic regression and a mathematical model is develop to study its relationship with faults. The results show that only Depth of Inheritance Tree (DIT) metrics is useful in fault-proneness prediction in the open source applications.