Authors: Matt Martin, Mitch Rees-Jones
This repository contains data science experiments for predicting time required to close issue reports ("issue close time").
There are 10 issue lifetime datasets located in data/. They were extracted from 10 large open source software projects by Matt Martin and used to build prediction classifiers. This project uses the time-independent features from Kikas, Dumas, and Pfahl's 2016 paper on predicting issue close time, as described in the following table:
| Feature name | Feature Description |
|---|---|
| issueCleanedBodyLen | The number of words in the issue title and description. For JIRA issues, this is the number of words in the issue description and summary |
| nCommitsByCreator | Number of commits made by the creator of the issue in the 3 months before the issue was created |
| nCommitsInProject | Number of commits made in the project in the 3 months before the issue was created |
| nIssuesByCreator | Number of issues opened by the issue creator in the 3 months before the issue was opened |
| nIssuesByCreatorClosed | Number of issues opened by the issue creator that were closed in the 3 months before the issue was opened |
| nIssuesCreatedInProject | Number of issues opened in the project in the 3 months before the issue was opened |
| nIssuesCreatedInProjectClosed | Number of issues in the project opened and closed in the 3 months before the issue was opened |
| timeOpen {1,7,14,30, 90,180,365,1000} | Close time of the issue. For example, |
Compile the Java classes:
$ make (or "make compile-java")
Configure the experimental setup by changing the variables at the top of run.sh
Run the experiment:
$ bash run.sh
Results can be found in the out/ directory.
Compile the Java classes:
$ make (or "make compile-java")
- Run the round robin experiment:
$ bash roundRobin.sh
Results can be found in the out/roundRobin directory.