A Tree-Based Classification of High and Low Error-Severity Modules Using the CK Metrics

摘要

Few research papers have addressed classifying modules into several levels of error severity using tree-based techniques. This research aims to provide a method of error classification at two levels: classifying classes into either erroneous or not-erroneous using OO metrics (binary category), and classifying classes into two severities of errors (Low or High) using OO metrics (multi-category). This paper analyzes an open-source system-Eclipse IDE-which was developed using the object-oriented paradigm. The Chidamber and Kemerer metrics are used to build a classifier of erroneous modules. While using the binary category classifier gives good results, we found that using the multi-category classifier produces a better classification performance.