A Knowledge Graph–Based Framework for Predicting Sprint Performance in Agile Software Projects

Agile software development is widely adopted in modern software organizations because it enables teams to deliver software incrementally and respond quickly to changing requirements. Despite its advantages, analyzing sprint performance remains challenging since project data are often distributed across multiple tools and systems and do not clearly represent the relationships between development activities. This study examines factors influencing sprint performance using knowledge graph modeling combined with regression analysis. Project data describing sprints, tasks, team members, and requirement changes were collected from agile development records. After preprocessing, these data were organized into a knowledge graph using the Neo4j graph database. This structure makes it possible to examine relationships such as task dependencies and coordination among team members during sprint execution. The study focuses on indicators related to task dependencies, workload distribution, and coordination patterns. The analysis is based on data from 380 sprints carried out by six agile development teams over a five-year period. Sprint performance is evaluated using the Sprint Completion Rate, which indicates how much of the planned story point workload is completed within a sprint. Task dependency depth is related to sprint performance, with higher values associated with improved sprint completion in this dataset. In addition, higher betweenness centrality (normalized) within the task dependency network is positively related to sprint performance, suggesting that well-coordinated task structures contribute to more effective sprint execution. The regression model explains approximately 48.9% of the variation in sprint completion rates. Overall, the results suggest that structural indicators derived from knowledge graphs help reveal patterns in agile development processes that are difficult to observe using traditional tabular data. The integration of graph-based indicators with regression- based analysis provides a practical framework for understanding sprint performance and supporting decision-making.