Optimization of the Bugs Classification of the Ticketing System in Software Development: a Study Case

Computer bug elimination is an important phase in the software development process. A ticketing system is usually used to classify the identified bug type and to assign a suitable developer. This system is handled manually and error prone. This paper proposes a new bug classification method using the fast string search algorithm. The method searches the error string and compares it to the full text. The approach is deployed to the software development process at PT. Selaras Anugerah Lestari and it results in a significant reduction in the average value of the time required to handle the bugs.


I. INTRODUCTION
PT. Selaras Anugerah Lestari is a newly established company.Its main business is the products related to hospitality.It sales hospital care unit and supporting tools.In order to run the business smoothly, it develops a number of applications such as Credit Approval Memorandum (CAM) system, Internal Memo, IT asset system, and Financial and HR system.It uses the ticketing system to record the bugs that occur during the development process.The ticketing is also used for the bug classification and to notify the relevant developers.Currently, the classification process is performed manually and it has the potential of human errors especially in the cases where the application contains many bugs.
A series of research in the area of software bug classification has been initiated since a couple of years ago [1][2][3][4][5][6][7][8][9][10][11][12][13][14].The first research was done by Ref. [15].They introduced a new classification system using a supervised text categorization so-called naïve Bayes (NB).The result showed that the accuracy level was about 30%.The accuracy level is a condition where the identified bug type matches the skill sets of the assigned developer.The other research were performed by Refs.[16,17].They used the earlier research as the basis for their research and added a number of extra features.Theirs approach is also able to provide a list of the recommendation of the developers.The results showed that the accuracy level had increased to 57% for the Eclipse software project and 74% for the Mozilla software project.
In this paper, we propose an approach to improve the process of bug classification using Fast String Searching Algorithm.Reference [18] stated that the Boyer Moore algorithm is extremely efficient in most cases in comparison to Knuth-Morris-Pratt and Brute Force algorithm.Reference [19] compared various exact string matching algorithms for virus signaturedetection and they concluded that the best algorithm for a common searching purpose is Boyer Moore.The Boyer Moore algorithm has no special memory requirements and needs no preprocessing or complex coding and thus can be surprisingly fast.
Every bug consists of two important parts: the bug's description and the application's name.The bug description expresses the detail information about the error that happens in the application.Meanwhile, the application name is the application where the bug happens.Figure 1 shows the example of the bug description and the application name.Figure 1 shows all the important part of the bug which is the bug description and application name.The information is text based; thus, they perfectly match with the capacity of the current algorithm.

II. RESEARCH METHOD
The idea of the proposed method is to change the manual classification process with the newly automatic one using fast search string algorithm.The fast search string algorithm is an algorithm which uses the searching string and compares it to the full text.Figure 2 shows the algorithm.Figure 3 shows an example of the fast string search algorithm where it is used to search the bug on the application name 'CAM'.Once the error is found, an appropriate software developed can assigned accordingly.

III. RESULTS AND DISCUSSION
The results of the study are depicted in Tables I and II.The former table shows the required time for bug classification without and with fast search string classification algorithm.From the table, one can see that the algorithm generally requires shorter bug classification time.
The difference between the two classification time is statistically evaluated using the paired t-test for comparison of the mean of two samples.Table II compares a number of statistics of the two samples.One can say that on average, the algorithm is able to reduce the classification time.The Pearson correlation suggests that the reduction occurs uniformly across the variation of the bug number.

IV. CONCLUSION
Software bugs are often unavoidable in the software development process.Many software developers utilize a ticketing system for bug classifications to facilitate an efficient bug elimination process.This research proposes the use of the fast string search algorithm for the bug classification.The proposed method is evaluated empirically.The results suggest that the new approach is able to reduce the time required for the bug classification.

Fig. 1 .
Fig. 1.The Example of the Bug's Description and Application Name.Stringlen  Length of String I  patlen Top: if I > stringlen then return false J < pathlen Loop: if j = 0 then return j=+1 if string(i) = pat(i) then J  j -1 I  i -1 go to Loop i  i + max(delta1, (string(i)), delta2, (j)) go to Top Fig. 2. The Fast String Search Algorithm.
Fig. 3.An Example of the Process of Fast String Search Algorithm.

TABLE I THE
TEST RESULTS: BUG HANDLING TIME WITH AND WITHOUT OPTIMIZATION.Cite this article as: Danar Ardhito and Abba Suganda Girsang, "Optimization of the Bugs Classification of the Ticketing System in Software Development", CommIT (Communication & Information Technology) Journal 10(2), 85-87, 2016.