Analysis Of Change Request Report And Research Of Prediction For Issue Tracking System

In order to measure and predict the closed-probability of a single change request of Issue Tracking System,several metrics are defined to measure the feature of change request,and built Logistic Regression prediction model using these best predictive metrics on datasets for training.Then experiments which applied pro-posed method were performed on datasets of testing which contain 20 SourceForge projects,and achieved a result that average recall of 95%and average FPR(False Positive Rate)of 14%.Also we get a result that average recall of 96.5%and FPR of 56.1%when we transfer the five metrics on SourceForge to JIRA projects.Analyses of experimental result show that the proposed method achieved a good prediction performance on test datasets of different Issue Tracking System,and closed-percentage or size of change requests report doesn't affect the performance of the model,and some features of change request report can be used to predict its closed-probability in the next version.The change requests on Issue Tracking System can produce a Associated Network of Change Requests(ANCF)because the dependence,impact between reports.Detection and identification of the most important change requests in network can help developers to better understand,implements,fix the requirement that change requests refers.We crawled the "Issue Links" data of Apache Hadoop and its related project.We build Associated Network of Change Requests on 54 projects which include 46759 pairs of associated relations.Then we analysis the importance of change requests in the network using 5 metrics.And we conduct correlation verification using the result of closed-probability prediction of a single change request and the measure metrics of Associated Network of Change Requests.We achieve a result that relevance and significance of closed-probability and local measure metrics(out degree and degree)is better than global and random walk metrics,but the correlated degree value is weak.The experimental result shows that these change requests is more complicated which includes "Issue Links" when we analysis their priority.We need fully consider the individual and its position in the global Associated Network of Change Requests.We must pay more attention to these backward associated requirements which contains "Issue Links",the more of backward requirement,the more analysis by synthesis.