A classification scheme for studies on fault-prone components
Author
Summary, in English
Various approaches are presented in the literature to identify fault-prone components. The approaches represent a wide range of characteristics and capabilities, but they are not comparable, since different aspects are compared and different data sets are used. In order to enable a consistent and fair comparison, we propose a classification scheme, with two parts, 1) a characterisation scheme which captures information on input, output and model characteristics, and 2) an evaluation scheme which is designed for comparing different models' capabilities. The schemes and the rationale for the elements of the schemes are presented in the paper. Important capabilities to evaluate when comparing different models are rate of misclassification, classification efficiency and total classification cost. Further, the schemes are applied in an example study to illustrate the use of the schemes. It is expected that applying these schemes would help researchers to compare different approaches and thereby enable building of a more consistent knowledge base in software engineering. In addition it is expected to help practitioners to choose a suitable prediction approach for a specific environment by filling out the characterisation scheme and making an evaluation in their own environment.
Publishing year
2001
Language
English
Pages
341-355
Publication/Series
Lecture Notes in Computer Science
Volume
2188
Document type
Conference paper
Publisher
Springer
Topic
- Computer Science
Status
Published