A systematic mapping study on cross-project defect prediction

TL;DR

This paper systematically reviews 50 studies on cross-project defect prediction, highlighting the heterogeneity in methodologies and the difficulty in comparing results across different approaches.

Contribution

It provides a comprehensive mapping of existing research on CPDP, identifying methodological differences and gaps in comparative analysis.

Findings

High heterogeneity in case study setups

Difficulty in qualitative comparison of approaches

Lack of consensus on best-performing methods

Abstract

Cross-Project-Defect Prediction as a sub-topic of defect prediction in general has become a popular topic in research. In this article, we present a systematic mapping study with the focus on CPDP, for which we found 50 publications. We summarize the approaches presented by each publication and discuss the case study setups and results. We discovered a great amount of heterogeneity in the way case studies are conducted, because of differences in the data sets, classifiers, performance metrics, and baseline comparisons used. Due to this, we could not compare the results of our review on a qualitative basis, i.e., determine which approaches perform best for CPDP.