Applying FAHP to Improve the Performance Evaluation Reliability and Validity of Software Defect Classifiers
Loading...
Authors
Ghunaim, Hussam
Issue Date
2019-10
Type
Thesis
Language
en_US
Keywords
Computer science , Data mining , Fuzzy analytical hierarchy process , Fuzzy analysis , Software defects
Alternative Title
Abstract
Today’s Software complexity makes developing defect-free software almost impossible. On an average, billions of dollars are lost every year because of software defects in the United States alone, while the global loss is much higher. Consequently, developing classifiers to classify software modules into defective and non-defective before software releases, has attracted a great interest in academia and the software industry alike. Although many classifiers have been proposed, none has been proven superior to others. The major reason is that while a research shows that classifier-A is better than classifier-B, we can find other research coming to a diametrically opposite conclusion. These conflicts are usually triggered when researchers report results using their preferred performance quality measures such as recall and precision. Although this approach is valid, it does not examine all possible facets of classifiers’ performance characteristics. Thus, performance evaluation might improve or deteriorate if researchers choose other performance measures. As a result, software developers usually struggle to select the most suitable classifier to use in their projects. The goal of this dissertation is to apply the Fuzzy Analytical Hierarchy Process (FAHP) as a popular multi-criteria decision-making technique to overcome these inconsistencies in research outcomes. This evaluation framework incorporates a wider spectrum of performance measures to evaluate classifiers’ performance, rather than relying on selected, preferred measures. The results show that this approach will increase software developers’ confidence in research outcomes, help them in avoiding false conclusions and indicate reasonable boundaries for them. We utilized 22 popular performance measures and 11 software defect classifiers. The analysis was carried out using KNIME data mining platform and 12 software defect data sets provided by NASA Metrics Data Program (MDP) repository.
Description
Citation
H. Ghunaim, "Applying FAHP to Improve the Performance Evaluation Reliability and Validity of Software Defect Classifiers", Ph.D. dissertation, Dept. of Engineering, Univ. of Bridgeport, Bridgeport, CT, 2019.