Fuzzy Delphi Method for Evaluating HyTEE Model (Hybrid Software Change Management Tool with Test Effort Estimation)

When changes are made to a software system during development and maintenance, they need to be tested again i.e. regression test to ensure that changes behave as intended and have not impacted the software quality. This research will produce an automated tool that can help the software manager or a maintainer to search for the coverage artifact before and after a change request. Software quality engineer can determine the test coverage from new changes which can support cost estimation, effort, and schedule estimation. Therefore, this study is intended to look at the views and consensus of the experts on the elements in the proposed model by benefitting the Fuzzy Delphi Method. Through purposive sampling, a total of 12 experts from academic and industrial have participated in the verification of items through 5-point linguistic scales of the questionnaire instrument. Outcome studies show 90% of elements in the proposed model consists of change management, traceability support, test effort estimation support, regression testing support, report and GUI meet, the value threshold (d construct) is less than 0.2 and the percentage of the expert group is above 75%. It is shown that elements of all the items contained in the venue are needed in the HyTEE Model (Hybrid Software Change Management Tool with Test Effort Estimation) based on the consensus of experts. Keywords—Fuzzy Delphi Method; software traceability; test effort estimation; regression testing; software changes


I. INTRODUCTION
The software application is present in every area of our life. The small and large system is developed using the software. Change is a part of everyday life. Software changes after some time. In today's competitive atmosphere, brand new needs are arising, and existing needs are altering swiftly. Changes are accomplished for various reasons, for example, to include new elements, to amend a few errors or to improve the product.
According to Vasa [1], "Software evolution or changes are direct consequences and reflections of ongoing changes in a dynamic real world". These changes are occurring very fast because of the competitive market. Enhancing software is a common necessity in business today as they encounter lots of need changes, prolonging software features and function, and including brand new modules. We cannot ignore the critically of software changes because real software system changes and becomes more complex over time [1]. A current report distributed by the Standish Group International [2] which involved 13522 software projects, Fig. 1 showed that out of the reviewed projects just 29 percent were effective, 18 percent is considered as "failed" and 53 percent are viewed as "suspected" and the fundamental driver of the failed project is the prerequisite change. Lam [3] propose that changing necessity are the main issues of the re-building and maintenance activities. The majority of the previous study demonstrates that software maintenance activities are concerning adaptive and completeness maintenance close to 80%. For this aim, the company must get the opportunity to manage requirement adaptation as part of the border software evolution approach. One estimate expresses that 40% of the necessity requirement during software development [4].
Estimation is limited as the shrewd conviction of the quantum or field that should be performed and the essential material (in particular, HR, money related assets, material assets, and time assets) required playing out the work at a future date in a characterized domain for determined strategies. Test Effort Estimation that estimate of the testing length, exertion, cost and timetable for a specific programming test project in an individual domain for particular strategies, tool and methods [6]. The test effort is the foundation of the effort spent on test action and the effort spent on debug action [7].
Research objective for this study is: 1) To develop a software traceability model to reduce operational cost during regression testing using the Fuzzy Delphi Method.  This study utilizes the FDM as an evaluation to evaluate the proposed model. This paper is organized as follows. Section I is an Introduction to the background of the problem, Section II introduces the process of the FDM to assess the expert consensus and lists the alternative options in the order of preference. Section III shows results and finding of an element in Hytee model. The conclusion of the research findings and future work of this research is presented in Section IV. The results are expected to provide an element to support in design and development of Hytee Model.

A. Introduction
This research study is about implementing the Fuzzy Delphi Method in designing and developing a software traceability model with the test effort estimation during regression testing in software changes. The Delphi Method is an approach that has been used and widely accepted to collect data for a study based on the validation expert in the research study of Hsu [8]. The strength of this method has also produced a diversity technique in obtaining empirical data like the Fuzzy Delphi Method (FDM). Talking about FDM, it is a method of measurement based on the modification on the Delphi Method.
This method has been presented by Kaufman and Gupta in 1988 [9]. The Fuzzy Delphi Method (FDM) is a combination of the numbering of the fuzzy set method and Delphi itself [10]. This brings the meaning that this is not a new approach based on a classical Delphi method where the respondents involved must be from within the circle of experts who have experience in the context of the study. This improvement indirectly strives to make FDM as a measurement approach that is more effective, whereby FDM is able to resolve the issue of who has uncertainty for some issues of the research.
The review of previous literature shows that FDM is a combination of the traditional method of Delphi (Classic) and fuzzy set theory (Fuzzy). The fuzzy set theory was introduced by an expert in the field of mathematics in 1965 which Zadeh [11] worked, and it works as an extension of a classical set theory where each element in a set is assessed based on the set of binaries (Yes or No). Fuzzy set theory assessment also allows a gradual review of each element. Ragin [12] states that the value of numbering fuzzy consists of 0 to 1 or in the unit interval (0,1).
There are two mains in FDM which is Triangular Fuzzy Number and Defuzzification Process. Triangular Fuzzy Number is m is made up of the value of the m1, m2, and m3 where m1 represents the value of the minimum (smallest value), representing the most reasonable value m2 (most plausible value) and m3 is referring to the maximum value (but there is value). All three values in the Triangular Fuzzy Number this can be seen through Fig. 2 shows the graph that triangles mean against the value of triangular shows that all three of these values is also in the range of 0 to 1 and it coincided with fuzzy numbers [12].

B. Procedure in FDM
For further details on the findings using the Fuzzy Delphi approach method (FDM), there are procedures that must be compiled. Table I show about the procedures in FDM cover for seven steps.

C. The Number of Expect in FDM
The selection of the number of experts for the Fuzzy Delphi Method (FDM) is a total of 12 people. This is based on the view of Adler and Ziglio [13], who pointed out that the number of experts for the Delphi technique was as many as 10 to 15 people if the experts can get an agreement with each other. However, there is also an opinion stating that the minimum number of experts for the Delphi technique is five experts [14]. This matches the argument from Rowe and Wright [15] that the number of experts can start from 5 to 20 people based on their areas of expertise. On the other hand, Jones and Twiss [16] suggested the number of experts involved in the Delphi method approach is 10 to 50 experts.  The criteria and characteristics of this study involved software engineering specialists in software testing from the academic and industrial sector. Selection is also based on Berliner [17], [18] who argues that the expert is competent if they participated in a particular field consistently exceeding a period of 5 years. Nonetheless, there are other scholars pointing out that experts are highly skilled and experienced in the areas studied [19] [20].
Based on Table II, the discussion for selection of the respondents for the design and development phase, the researcher lists down the criteria for selecting the experts, as below:

D. Questionnaire
This research uses the questionnaire as an instrument to get quantitative data for element requirements of Hytee Model. The questionnaire is aimed at in order to meet the criteria and conditions of the using the technique of fuzzy Delphi Method where this technique involves the use of a mathematical formula in order to obtain the consensus of experts. Instruments used by researchers is the instrument that has been modified based on the needs of the study researchers. The original of this questionnaire was adapted from the study of Ibrahim (2006) [21]. Table III shows the element in the questionnaire.
The process of data collection in the study is carried out using the Fuzzy Delphi approach between the processes involved in an interview for the Delphi technique while the questionnaire is analyzed with techniques of a fuzzy number. 5-point scale used Rahman, M. N. A. (2013) [23] to determine the expected kinds of video games against aspects of basic skills in the Malay language for foreign students to obtain consent or consensus of a group of expert. To facilitate experts, answer questionnaires, researchers have put the value of the scale of 1 to 5 to replace the Fuzzy value as shown in Table IV for linguistic scale 5 points follows;

III. RESULT AND FINDING
Data analysis is to follow the approach of Fuzzy\ Delphi through step 3 to 7 will answer questions the study disclosed. For viewing the degree of agreement among experts, the findings for each of the items were analyzed by a Threshold value (d) for two fuzzy numbers m = (m1, m2, m3) and n = (m1, m2, m3) are calculated using the formula: It is supported by the Rahman, 2013 [22] and Jamil, 2013 [23] which States that in order to analyze the data, the distance between two Fuzzy number is calculated by measuring the average value of the deviation between the experts. Whereas the criteria used to assess the expert group consensus is based on the degree of agreement in excess of 75%.
In this study, one (1) is complied with because the value threshold for most of the subitem is ≤ 0.2, but only at part subitem only. However, the second condition (2) has also been observed because the expert group consensus is above 75%. Result value threshold ≤ 0.2, indicating that this study gets the value of the threshold exceeds 75% 77.8% by registering for a theme that includes a total of 5 subitems. This shows the degree of agreement among the experts has reached a consensus that good. Therefore, the second round for fuzzy Delphi is not needed because of data acquisition complies with both conditions. Below show the findings for elements of Hytee model based on the consensus of experts. This data consists of the value of the threshold each element (d item), the value threshold constructs (d)

D. Test Effort Estimation Support
Table VIII display findings for Test Effort Estimation Support components for Hytee proposed model on the consensus of experts using the Fuzzy Delphi Method (FDM). The findings of this study show the value threshold (d) and the percentage of the expert group. From the findings, all the item meets the expert consensus > 75.0 %. The first two elements, in change management and the report, has an average value of "d" threshold of less than 0.2. Accordingly, both have reached the percentage of expert consensus of more than 75%, and the defuzzification scores greater than 0.5, making them acceptable as antecedents for the customer engagement concept studied. This study for evaluation Hytee Model 2 item needs to revise again and update in the Hytee System. That action was done to able the whole % item "d" ≤ 0.2 has achieved the agreement of 78%, making this construct successfully maintained. It has concluded that all the elements in Hytee Model (except 2 elements) are maintained and certified of Hytee model based on the consensus of an expert. Using Fuzzy Delphi Method analysis, this study has proven the importance of element in Hytee Model. The findings of the study are in line with its purpose to answer the questions pertaining to the agreement of experts on an element into work developing proposed Hytee Model. The defuzzification process is greatly used to filter the priority of element. In change management, the contribution of this result to ensure the user understands the environment of change management. For traceability support the contribution of this study to prove the flow of the system using the traceability model. For regression testing support refer to Table VII is to show the function of regression testing in Hytee Model. In test effort estimation refer to Table VI show the user the result after all the flow of change of the error. For report and GUI refer to the table, we can see that the expert agree with the GUI of Hytee Model.
As future work, in this stage, the researcher will design the model based on the data from the Fuzzy Delphi Method discussion within the expert review. From elements of change management, traceability support, regression testing support, test effort estimation support, report and GUI, the findings it to continue to upgrade the proposed model to the actual model for improvement.