Home // SOFTENG 2024, The Tenth International Conference on Advances and Trends in Software Engineering // View article

Software Bug Prediction Based on Semi-definite Logistic Regression Model

Authors:
Tadashi Dohi
Jingchi Wu
Hiroyuki Okamura

Keywords: software bug prediction; bug-prone module; logistic regressions; semi-definite programming; discrimination problem; F-score.

Abstract:
In software bug prediction to identify bug-prone modules, several machine learning techniques have been used in past. However, it has been known that almost all of them were not explainable and could not be applied to the program understanding, because the contributions of software metrics were unclear in such black box techniques. In this article, we aim at overcoming the problems in an explainable logistic regression model, called multicollinearity and interaction, and apply the semi-definite logistic regression model to identify software bugprone modules. More specifically, we use three actual software development project data sets to evaluate the F-score, as wellas precision and recall, and compare our semi-definite logistic regression model with the classical logistic one, in terms of the predictive performance of software bug-prone modules. It is shown that our semi-definite logistic regression model involves the common logistic regression model as a special case and can improve the predictive performances on the F-score.

Pages: 11 to 16

Copyright: Copyright (c) IARIA, 2024

Publication date: May 26, 2024

Published in: conference

ISSN: 2519-8394

ISBN: 978-1-68558-178-7

Location: Barcelona, Spain

Dates: from May 26, 2024 to May 30, 2024