Investigating the Effect of Object-oriented Metrics on Fault Proneness Using Empirical Analysis

This paper presents an innovative metric based on a class abstraction to capture aspects of software complexity through combinations of class characteristics. The study also used software metrics effectiveness in finding the classes in different error categories for the three versions of Eclipse, the Java-based open-source Integrated Development Environment. Many studies used Logistic regression models to investigate the ability of OO software metrics to predict fault prone classes. We also used this method not only for binary but also multinomial categorization and empirically validate the ability of metrics to predict fault prone classes in different category using fault data. We conclude that this proposed metric is as effective as the traditional metrics in identifying fault-prone classes in binary categorization and also showing most efficient result for multinomial categorization. We also find that Univariate model for these metrics have same performance as the individual metric with no any learning technique in prediction of fault-proneness.