Predicting Reassignments of Bug Reports — an Exploratory Investigation
1. Predic'ng
Reassignments
of
Bug
Reports
—
an
Exploratory
Inves'ga'on
Ahmed Lamkanfi, Serge Demeyer
Ansymo
Antwerp Systems and Software Modelling
Proceedings of the 17th European Conference on Software Maintenance and
Reengineering
Wednesday 12 June 2013
5. Provides vital information
Limited technical background
Report through the bug-tracking system
DevelopersEnd-users
Wednesday 12 June 2013
6. Provides vital information
Limited technical background
Report through the bug-tracking system
DevelopersEnd-users
Wednesday 12 June 2013
7. Provides vital information
Limited technical background
Report through the bug-tracking system
DevelopersEnd-users
Potential incorrect information
Wednesday 12 June 2013
8. Provides vital information
Limited technical background
Report through the bug-tracking system
DevelopersEnd-users
Component attribute
Potential incorrect information
Wednesday 12 June 2013
10. /10
Product JDT
Compone
nt
Core
Severity Normal
Summary
“Dangerous Eclipse
behavior on file save
when out of disk space”
Product Platform
Component Text
Severity Critical
Summary
“[implementation]
Dangerous
Eclipse behavior on file
save when out of disk
space”
3
Wednesday 12 June 2013
11. /10
Product No. of Bugs No. of Components
Platform 35.378 21
PDE 7.136 5
JDT 16.472 6
CDT 7.750 19
GEF 852 7
Core 54.999 116
Bugzilla 5.274 21
Firefox 67.348 28
Thunderbird 18.513 18
SeaMonkey 7.808 33
Cases
under
Study:
Eclipse
and
Mozilla
4
Wednesday 12 June 2013
12. /10
Product No. of Bugs No. of Components
Platform 35.378 21
PDE 7.136 5
JDT 16.472 6
CDT 7.750 19
GEF 852 7
Core 54.999 116
Bugzilla 5.274 21
Firefox 67.348 28
Thunderbird 18.513 18
SeaMonkey 7.808 33
Cases
under
Study:
Eclipse
and
Mozilla
4
Some products with
large number of components
Wednesday 12 June 2013
13. /10
Cases
under
Study:
Eclipse
and
Mozilla
Product No. of Reassigned % of Reassigned
Platform 8326 23.5
PDE 1290 18.1
JDT 4952 30.1
CDT 2328 30.0
GEF 71 8.3
Core 17999 32.7
Bugzilla 489 9.3
Firefox 10474 15.6
Thunderbird 1697 9.2
SeaMonkey 1224 15.7
5
Wednesday 12 June 2013
14. /10
Cases
under
Study:
Eclipse
and
Mozilla
Product No. of Reassigned % of Reassigned
Platform 8326 23.5
PDE 1290 18.1
JDT 4952 30.1
CDT 2328 30.0
GEF 71 8.3
Core 17999 32.7
Bugzilla 489 9.3
Firefox 10474 15.6
Thunderbird 1697 9.2
SeaMonkey 1224 15.7
5
✓
Many reassignments
✓Difference between Eclipse and Mozilla
Wednesday 12 June 2013
15. /10
1. Many reports are assigned to wrong component
6
Wednesday 12 June 2013
16. /10
1. Many reports are assigned to wrong component
2. More bug triaging steps needed
6
Wednesday 12 June 2013
17. /10
1. Many reports are assigned to wrong component
2. More bug triaging steps needed
3. Decrease of productivity
6
Wednesday 12 June 2013
18. /10
1. Many reports are assigned to wrong component
2. More bug triaging steps needed
3. Decrease of productivity✓ Tool support to aid bug triagers is useful
6
Wednesday 12 June 2013
19. /10
History of all reported bugs
Uncover/mine facts about history
Make predictions about future
Bug Database
7
Wednesday 12 June 2013
20. /10
Can we predict which bugs will be assigned to a
different component?
8
Wednesday 12 June 2013
21. /10
Can we predict which bugs will be assigned to a
different component?
Binary classification problem
8
Wednesday 12 June 2013
22. /10
Can we predict which bugs will be assigned to a
different component?
Binary classification problem
Summary and initial component as input
8
Wednesday 12 June 2013
23. /10
Can we predict which bugs will be assigned to a
different component?
Binary classification problem
Summary and initial component as input
Each product with separate classifier
8
Wednesday 12 June 2013
24. /10
Can we predict which bugs will be assigned to a
different component?
Binary classification problem
Summary and initial component as input
Precision and Recall of predictions
Each product with separate classifier
8
Wednesday 12 June 2013
25. /10
Can we predict which bugs will be assigned to a
different component?
Binary classification problem
Evaluation using 10-fold cross-validation
Summary and initial component as input
Precision and Recall of predictions
Each product with separate classifier
8
Wednesday 12 June 2013
30. /10
Conclusions
10
Bugs are often assigned to the wrong
component
Opportunity to improve bug triaging process
by predicting component reassignments
Wednesday 12 June 2013