Semantic Graph Analytics for Research Trends in Software Defect Prediction

Software defect prediction (SDP ) is a critical area in software engineering quality assurance, leveraging predictive analytics to enhance defect identification and resolution. However, existing approaches often narrowly focus on single methodologies or lack consensus-driven insights, limiting their ability to inspire future research. This study addresses these gaps by exploring semantic relationships among 31 published studies, including journal articles, conference proceedings, and book chapters, using a novel graph analytics framework. Research articles were represented as nodes, and edges were formed based on a minimum threshold of five shared semantic terms. Degree centrality was computed to quantify the influence and connectivity of articles within the research network. Results revealed a significant clustering of studies, with a conference proceeding exhibiting the highest degree centrality (0.309859), indicative of its methodological influence on 59.7% of the network. Conversely, 40.2% of articles were isolated, highlighting unique research directions, including sentiment analysis of defect reports. Findings provide critical insights into the consensus and divergence in SDP research, emphasizing the importance of ensemble methods, severity prediction, and feature selection while revealing gaps in methodological overlap and emerging topics. This work demonstrates the utility of semantic graph analytics in extracting valuable knowledge from scholarly literature to advance the SDP domain.