International Journal of Scholarly Research (IJSR)

Vol-1, Issue-1, 2017

http://www.scholars.org.in

Review paper on Phenomenon of bug triage system"

Sakshi D. Gabhane1, Asst.Prof. Miss. Rana Sayeda 2

1
Departement of computer science and engg,
Jhulelal Institute of Technology, Nagpur
sakshigabhane@gmail.com
2
Departement of computer science and engg,
Jhulelal Institute of Technology, Nagpur
raana.syeda@gmail.com

Abstract: Nowadays IT companies is spending more than 40 repository provides a data platform to support many types of
percent of their cost in fixing software bugs, traditionally these tasks on bugs, e.g., fault prediction, bug localization, and
bugs are fixed by manual assignment to a particular developer reopened bug analysis.
, this approach causes too much dependency, the new and The bug reports in a bug repository are called bug
alternative approach is the bug triage system which fix the bug data. This bug report is then assigns to a developer, who starts
automatically , which automatically assign the reported bug to to fix the bug. If the assigned developer is unable fix the bug,
a develop which decreases the time and cost in in manual the bug is migrated to another developer. The process of
work, different classification techniques are used to conduct assigning a bug report to an appropriate developer is called
automatic bug triage. In this paper, we propose to apply bug triage. As the Work of Bug triage system is to choose the
machine learning techniques to assist in bug triage to predict appropriate developer for fixing bugs, we will follow the
which developer should be assigned on the bug based on its existing work to remove unfixed bug reports.
description by applying text categorization. We will address
the problem of data reduction for bug triage, i.e. how the
quality of bug data would be improved.

Keyword: Bug triage, text categorization, bug data, data 2. RELATED WORK
reduction. Cubrani and Murphy First proposed the problem of automatic
bug triage. The Machine Learning technique, Text
Categorization is applied to assist in bug triage by using text
categorization [1]. Text categorization is also known as text
1. INTRODUCTION
classification which is a technique of automatically sorting a
Automated Bug tracking system has its significance in large set of documents into categories from a predefined set where a
software development projects which manages bug reports and developer gets predicted using the bugs description [2]. For
list of developers who work on xing them. Bug tracking this they used supervised machine learning technique using
systems has its importance in open source software Nave Bayes classifier to predict the correct developer.
development, where the team members can be dispersed Xuan present a semi-supervised approach for
around the world. In such distributed projects, the developers automatic bug triage using text classification [3]. Their
and other contributors may rarely see each other. Secondly, the approach combined the naive Bayes classification approach
bug tracking system is used not only to keep track of problem and expectation maximization to take the advantage of both
reports and feature requests, but it also help in coordinating the labelled and unlabelled bug reports. Xuan trains a classifier
work among the different developers. with a fraction of labelled bug reports. This approach labelled
Software bugs can never be avoidable and at the numerous unlabelled bug reports. From the result of, this semi-
same time fixing bugs is expensive in software development. supervised approach improves the classification accuracy of
IT companies spend more than 40 percent of their cost in bug triage by up to 6% and it avoids low-quality bugs.
fixing software bugs. Large software projects deploy bug When a bug report has been assigned to a specific
repositories to support information collection and to assist developer, then if the assigned developer is unable to fix the
developers to handle bugs. A bug repository will contain all bug, the assigned developers can forward or reassign the bug
report which plays an important role in managing software to other developer. This process of reassignment of bug from
bugs. In a bug repository, a bug is maintained as a bug report, one developer to another is called Bug Tossing. Jeong find
which records the complete description of reproducing the bug out that in manual bug triage, 37 percent - 44 percent of bug
and updates according to the status of bug fixing. A bug reports are tossed[4].In addition, Jeong et al. a model of bug

International Journal of Scholarly Research Page 6

International Journal of Scholarly Research (IJSR)
Vol-1, Issue-1, 2017
http://www.scholars.org.in

tossing is built to reduces the number of reassignment of bug done the real work on the report. This rule covers the
reports. The Markov chains based tossing graph approach is frequent case where an Eclipse developer files a
proposed to capture the past bug tossing history which report, which is then assigned to somebody else or a
improved the bug assignment and reduced unnecessary tossing sub-team alias by default, and then later implements
steps. the x himself.
P. Bhattacharya proposed a method for bug triage Scenario 4: If the report was resolved as non-xed
[12]. Goal of His Proposed system was to find the optimal set by the person who submitted it, and who was not
of machine learning techniques to improve bug assignment also assigned to it, the report is class labeled with the
accuracy in large projects. [5] Used a set of machine learning first assigned developer or person. Since it might be
tools and a probabilistic graph-based model (bug tossing a feature or a bug which will later be handled by
graphs) that lead to highly-accurate predictions, and laid the someone or after being prompted by a developer for
foundation for the next generation of machine learning-based details of his or her setup and discover that there is
bug assignment. They used methodology like Choosing no bug.
effective classifiers and features, Incremental learning, Multi Scenario 5: If the submitter resolve the report as
featured tossing graphs to achieve their goal. non- fixed who was not the assigned-to developer,
The current technique of Bug Triaging involves and nobody responded, there is an error or the
modelling the reassignment of bugs as a goal-oriented path submitter caught the mistake before anyone started
model [6]. V. Akila. Proposed a new framework with the to solve the issue, these reports are removed from
additional capabilities. This models the reassignment of bugs the training set, as it cannot be reliably labeled.
as Enriched Adaptive Bug Triaging System (EABTS) which is Scenario 6: If no developer could resolve the report
based on actual path model. Their graph structure captures the then it is marked as non-fixed and the class is
relationship among developers as the number of tosses and assigned by the developer who was the last person
also captures the propinquity exists among developers. who has worked on the report.
Therefore, this graph structure is enriched. The technique was
based on Ant routing. Ant routing is inherently adaptive in
nature. Their work gave a sub graph that consists of developers 4. CONCLUSIONS
who are frequently involved in bug resolution.
Bug triage is one of the expensive step in software
3. PROPOSED METHOD maintenance for both cost as well as time. Our Approach will
We propose to apply machine learning techniques to assist in combine feature selection with instance selection to reduce the
bug triage to predict which developer should be assigned on scale of bug data sets as well as improve the data quality. To
the bug based on its description by applying text determine the order of applying instance selection and feature
categorization, i.e. how the quality of bug data would be selection for a new bug data set, we will extract the attributes
improved. Following are the different scenarios for Our of each bug dataset and will train a predictive model based on
Proposed Bug Tracking System historical bug datasets. We will empirically investigate the
data reduction for bug triage in bug repositories of two large
Scenario 1: In first scenario the assigned developer open source projects, namely Eclipse and Mozilla. Our
will resolved the report and it will be labeled by the Proposed system will be based on Random Forest which will
developer's class regardless of who has submit the provide an efficient approach on data processing to reduce the
report or the type of resolution. scale and provide high-quality bug data in software
Scenario 2: In second scenario, the report can be development and maintenance.
resolved by someone other than the assigned
developer, but not by the person who submitted it or
nor by the person directly assigned, the report will be
labeled with the class of the particular developer who
marked it resolved. The reasoning is that whoever
resolves the report is the person to whom it should 5. REFERENCES
have been assigned.
Scenario 3: In this scenario, if the report is resolved [1] D. Cubrani and G. C. Murphy, Automatic bug triage
as fixed, Regardless of whom the resolver was, we using text categorization, in Proc. 16th Int. Conf. Softw.
assume that this is the developer who implemented Eng. Know l. Eng., Jun. 2004, pp. 9297.
the xed and label the report with the class of the
assigned developer, as He is the person who had
International Journal of Scholarly Research Page 7
International Journal of Scholarly Research (IJSR)
Vol-1, Issue-1, 2017
http://www.scholars.org.in

[2] J. Anvik, L. Hiew, and G. C. Murphy, Who should x

this bug? in Proc. 28th Int. Conf. Softw. Eng., May 2006, [14] S. Breu, R. Premraj, J. Sillito, and T. Zimmermann,
pp. 361370. Information needs in bug reports: Improving cooperation
between developers and users, in Proc. ACM Conf.
[3] J. Xuan, H. Jiang, Z. Ren, J. Yan, and Z. Luo, Automatic Comput. Supported Cooperative Work, Feb. 2010, pp.
bug triage using semi-supervised text classification, in 301310.
Proc. 22nd Int. Conf. Softw. Eng. Knowl. Eng., Jul. 2010,
pp. 209214 [15] V. Bolon-Canedo, N. S anchez-Marono, and A. Alonso-
Betanzos, A review of feature selection methods on
[4] G. Jeong, S. Kim, and T. Zimmermann, Improving bug synthetic data, Knowl. Inform. Syst., vol. 34, no. 3, pp.
triage with tossing graphs, in Proc. Joint Meeting 12th 483519, 2013.
Eur. Softw. Eng. Conf. 17th ACM SIGSOFT Symp.
Found. Softw. Eng., Aug. 2009, pp. 111120.

[5] T. Zhang, G. Yang, B. Lee, I. Shin Role Analysis-based

Automatic Bug Triage Algorithm, 2012.

[6] V.Akila, Dr.G.Zayaraz, Dr. V. Govindasamy Effective

Bug Triage A Framework, International Conference on
Intelligent Computing, Communication & Convergence,
114 120, 2015.

[7] S. Artzi, A. Kie _ zun, J. Dolby, F. Tip, D. Dig, A.

Paradkar, and M. D. Ernst, Finding bugs in web
applications using dynamic test generation and explicit-
state model checking, IEEE Softw., vol. 36, no. 4, pp.
474494, Jul./Aug. 2010.

[8] J. Anvik and G. C. Murphy, Reducing the effort of bug

report triage: Recommenders for development-oriented
decisions, ACM Trans. Soft. Eng. Methodol., vol. 20, no.
3, article 10, Aug. 2011.

[9] C. C. Aggarwal and P. Zhao, Towards graphical models

for text processing, Knowl. Inform. Syst., vol. 36, no. 1,
pp. 121, 2013.

[10] Bugzilla, (2014). [Online]. Avaialble: http://bugzilla.org/

[11] K. Balog, L. Azzopardi, and M. de Rijke, Formal

models for expert nding in enterprise corpora, in Proc.
29th Annu. Int. ACM SIGIR Conf. Res. Develop. Inform.
Retrieval, Aug. 2006, pp. 4350.

[12] P. S. Bishnu and V. Bhattacherjee, Software fault

prediction using quad tree-based k-means clustering
algorithm, IEEE Trans. Knowl. Data Eng., vol. 24, no. 6,
pp. 11461150, Jun. 2012.

[13] H. Brighton and C. Mellish, Advances in instance

selection for instance-based learning algorithms, Data
Mining Know l. Discovery, vol. 6, no. 2, pp. 153172,
Apr. 2002.

International Journal of Scholarly Research Page 8