Just-in-Time Software Defect Prediction Techniques: A Survey

Just-In-Time Software Defect Prediction (JIT-SDP) is a technique that allows teams to predict defects in software on the commit level instead of the release level. Using JIT-SDP will spare a lot of time and effort for teams, companies, and even clients. The need of predicting software defects in early phases urged researchers to study and propose many models in this direction. This paper aims to survey research in the last decade and inspect the evolution of JIT-SDP since the year 2012 until 2023. 30 research papers have been collected, classified, and discussed. It has been found that Machine Learning and Deep Learning are the most used techniques in this scope. Effort-Aware models and Cross-Project models are also two vital scenarios for JIT-SDP. Three research questions have been proposed in this research regarding the most frequent techniques used for prediction, the most frequent project (datasets), and the most frequent evaluation metrics adopted in the research. It has been intended to draw a timeline of the evolution of JIT-SDP models that can lead to more novel models to be proposed in the future.