Authors:
George Ciubotariu
;
Gabriela Czibula
;
Istvan Czibula
and
Ioana-Gabriela Chelaru
Affiliation:
Department of Computer Science, Babes,-Bolyai University, Cluj-Napoca, Romania
Keyword(s):
Machine Learning, One-Class Classification, Software Defect Prediction, Support Vector Machines.
Abstract:
Software defect prediction is a relevant task, that increasingly gains more interest as the programming industry expands. However, one of its difficulties consists in overcoming class imbalance issues, because most open-source software projects that are annotated using bug tracking systems do not have lots of defects. Therefore, the rarity of bugs may often cause machine learning models to dramatically underperform, even when diverse data augmentation or selection methods are applied. As a result, our focus shifts towards one-class classification, which is a family of outlier detection algorithms, designed to be trained on data instances of a single label. Considering this approach, we are adapting the traditional Support Vector Machine model to perform outlier detection. Experiments are performed on 16 versions of an open-source medium-sized software system, the Apache Calcite software. We are performing an extensive assessment of the ability of one-class classifiers trained on soft
ware defects to effectively discriminate between defective and non-defective software entities. The main findings of our study consist in uncovering several trends in the behaviour of the one- and binary-class support vector machine-based models when solving SDP problems.
(More)