Abstract:
Agile project management, focusing on flexibility and iterative delivery, has a tendency to view issue resolution timing as vague and thus has a tendency to result in schedule overruns and resource planning inefficiency. This research contributes a large-scale statistical approach to predicting issue resolution patterns from historical JIRA data, providing quantitative analysis to project managers to enhance sprint planning and resource allocation. It utilises time-series analysis with a broad dataset of 1,095 days of issue tracking records sourced from a publicly available Kaggle dataset of actual software development projects. It uses an AutoRegressive Integrated Moving Average (ARIMA) model, selected through a Box–Jenkins model-identification process that effectively tests various setups of parameters using stringent statistical tests. The optimal ARIMA (1,1,1) model demonstrated strong forecasting capability, as indicated by the performance metrics: AIC (Akaike Information Criterion) = 1872.52, BIC (Bayesian Information Criterion) = 1887.07, and error metrics recomputed on a single, consistent scale (Mean Absolute Error = [author to insert corrected value], Root Mean Squared Error = 3.2501; see Section 3. viii for the corrected computation). The results validate the efficiency of the model for forecasting issues resolved per day and establishing patterns over time in team productivity. This research provides robust variations in resolution rate prediction across development cycle stages, with high performance in stable sprints. The model output shows good estimation of team capacity for data-driven sprint backlogs and deadline realignment. This research adds a practical, scalable approach for JIRA-using teams, bridging the gap between Agile principles and data science.
