A Comparative Study of Bug Classification Algorithms

作者：Nagwani Naresh Kumar^*; Verma Shrish

来源：International Journal of Software Engineering and Knowledge Engineering, 2014, 24(1): 111-138.

DOI：10.1142/S0218194014500053

• Bug classification
• comparative study
• classifier performance evaluation
• classifier accuracy

The performance of ten classic algorithms to classify the software bugs for different bug repositories are compared. The algorithms included in the study are Naive Bayes, Naive Bayes Multinomial, Discriminative Multinomial Naive Bayes (DMNB), J48, Support Vector Machine, Radial Basis RBF) Neural Network, Classification using Clustering, Classification using Regression, Adaptive Boosting (AdaBoost) and Bagging. These algorithms are applied on four open source bug repositories namely Android, JBoss-Seam, Mozilla and MySql. The classification is evaluated using 10-fold cross validation technique. The accuracy and F-measure parameters are compared for all of the algorithms. The concept of software bug taxonomy hierarchy is also introduced with eleven standard bug categories (classes). The comparative study also covers the effect of number of categories over performance of classifiers in terms of accuracy and F-measure. The results are produced in tabular and graphical forms.

• 出版日期2014-2