Evaluating the Maintainability of a Software System by using Fuzzy Logic Approach

Published Online December 2014 in MECS DOI: 10.5815/ijitcs.2015.01.08

Quality of any software depends on the quality factors.

Among all the quality factors, software maintainability is broadly accepted as a highly significant contributor in the economic success of software systems and products. There is a demand for software engineers to understand and urge how various components of a software system is required in order to enhance the reliability of code during maintenance. Maintenance of software is the most expensive and resource requiring phase of the software development process. Statistics from various organizations shows that 40% to 80% of the development expenditure on the average software is spent in the maintenance phase [1, 2]. Maintainability evaluation is an essential component of modern software development life cycle. Maintainability evaluation if done correctly can be useful in software selection and costs estimation among different software, allocation of resource and staff, and so on. This minimizes the future maintenance effort [3].

According to ISO/IEC-9126, maintainability is the capability of the software to be modified [14]. These modifications can be corrective, adaptive or perfective in order for software to command with new requirements and functional specifications.

The ease with which the maintenance of a functional unit can be performed in accordance with prescribed requirements is all about maintainability. There are some major factors which controls maintainability.

The set of maintainability`s sub-characteristics can be classified in a hierarchical tree which consists of the characteristic of maintainability, its sub-characteristics and metrics. Maintainability is on the tree`s highest level while metrics are on the lowest. Maintainability is analyzed by in sub- characteristics which in turn can be evaluated by using metrics.

For evaluating the maintainability of object oriented system, four factors are taken i.e. complexity, coupling, inheritance and message passing. These factors are chosen since they are the design complexity factors and show more impact on the maintainability of object-oriented software system.

These factors are mapped with associated metrics for which Chidamber and Kemerer (CK) software metrics [15] have been used for maintainability evaluation. These metrics are focused on assessing the design of object-oriented system rather than implementation which makes them more suited to object-oriented paradigm as object-oriented design [16].

A case study is taken for the purpose. The complete evaluation of maintainability is done by using AHP and fuzzy methodology. This paper has designed in various sections.

Section 2 focuses on literature survey related with maintainability evaluation.

Section 3 discussed on Fuzzy Maintainability Model and their mentioned factors. Here primary goal is to present ISO/IEC 9126 model`s dimensions and their associate metrics.

Section 4 defines methodology of evaluation. It defines the software and programming language i.e. Java. This software majorly separates the modules of a management system and then membership mapping has been done. At last defuzzification of the comprehensive score has been done. Here AHP technique has played major role. Basically it’s a stepwise evaluation which holds technique, selected software and defuzzification.

• II.    Literature Survey

on Software maintainability metric by fuzzy integral. This paper proposed evaluation model of software maintainability and software maintainability index system. Kiranjit Kaur [13] worked on Multivariate linear model which estimated the maintainability of a class diagram in the term of reliability and portability. Metrics were chosen as CK and MOOD that may help software designer for improving the maintainability of a class diagram in design phase.

• III.    Fuzzy Maintainability Model

  Maintenance is a vital part of development with required and updated special techniques. Most often maintainability is performed without requirements or design documents or it may need basic code as well. Therefore these models are made to perform whole process orderly and efficiently In order to deal with real time applications with maintenance below is the proposed Fuzzy Maintainability model. The purpose of this study is to propose a evaluation model based on fuzzy evaluation approach and using the dimensions of ISO/IEC 9126.

Fig 1. Fuzzy Maintainability Model

This model is showing the metrics that are associated with the sub characteristics (Analyzability, Changeability, Stability and Testability) of Maintainability and will be evaluated further in terms of their metrics that are shown in Fig 1. Scope of studying this Fuzzy Maintainability model is the ease of evaluating Maintainability in Object oriented systems. The further results will be fuzzified. The model accesses modules comprehensively with defuzzified score. Brief outlines of all these factors are shown below [19, 20]:

Complexity defines the difficulty to preserve, modify and comprehend the software. Increased software complexity means that maintenance and enhancement projects will take longer, will cost more, and will result in more errors [21]. It is one of the parts in software metrics that is focused on direct measurement of software attributes, as opposed to indirect software measures such as project milestone status and reported system failures.

Coupling defines the interdependency between components or functions. Coupling is the measure of interconnections among the modules in a software structure. A the degree to which each program module depends on the other and it is required to achieve low coupling in software systems[22]. Coupling is a measure of interdependence of two objects. CBO for a class is measured by counting the number of other classes to which it is coupled. Two classes are coupled if methods of one use methods and/or instance variables of the other. High CBO indicates complex design, decreases modularity, and complicates testing of the class [17].

Inheritance is defined as classes having same methods and operations based on hierarchy. It is a mechanism whereby one object acquires the characteristics from one or more other objects. It is the length of the longest path from a given class to the root class in the inheritance hierarchy and is measured by the number of ancestor classes. So this metric calculates how a class is declared in the inheritance hierarchy. High DIT indicates greater design complexity and more fault-proneness [23].

Message Passing processes or objects can send and receive messages (having zero or more bytes, complex data structures or even segments of code) to other objects or processes. Processes can be synchronized by waiting for messages. It is used for the objects to gather them for a program to work. For objects and systems located on different computers, this is the source for interaction. Message Passing reduces the amount of coding logic and increases maintainability.

In the mentioned proposed model CK metrics are considered. CK metrics are prior when it comes to maintainability evaluation, as their metrics can be used to analyze Coupling, Cohesion, Size very well.

• IV.    Methodology

To evaluate maintainability, project on student management system developed in Java [34, 35] is considered. Tools selected as Analyst4j tool. Analyst4j is an Eclipse platform based tool which is available as a stand-alone Client Application or as an Eclipse IDE plugin. It features search, metrics, analyzing quality, and report generation for Java programs. Analyst4j tool are in demand for finding out the quality related metrics [36]. It generates various reports like shown in Fig 2. The AHP method is applied for judgment procedure [17, 18].

Table 1. Matrix representation for Eigen Vector

	COMP	COUP	INH	MSGP	The root	Eigen Vector(A)
COMP	1	1	1/3	5	1.136	0.226
COUP	1	1	1	3	1.316	0.262
INH	3	1	1	9	2.279	0.454
MSGP	1/5	1/3	1/9	1	0.293	0.058
Total					5.024	1.000

Where COMP-Complexity, COUP-Coupling, INH-

Inheritance and MSGP-Message Passing.

λmax =4.17, C.I = 0. 056, C.R =0.062

Therefore,

A = [0.226 0.262 0.454 0.058]

(i )

• A.    Determining the Fuzzy Member Function for Appraisal Matrix R

The following table 2 describes the fuzzy member function. This comprises metric ranges and divided on the scale of their performance .Eg Complexity varies from 020. Its associated metric will judge their category (Very poor, Poor, Medium, Good and Excellent) according to their ranges. Similarly, same goes for rest of metrics.

has 16 modules i.e.

Table 2. The membership mapping for metric score ranking

	Metric	Very Poor	Poor	Medium	Good	Excellent
COMP(0-20)	WMC	0<= x <5	5<= x <10	10<= x <15	15<= x <20	20<= x
COUP(0-5)	CBO	0<=x <1.25	1.25<=x<2.5	2.5<= x <3.75	3.75<= x <5	5 <= x
INH(0-3)	DIT	0<=x <0.75	0.75<= x <1.5	1.5<= x <2.25	2.25<= x <3	3<= x
MSGP(0-1)	RFC	0<=x <0.25	0.25<= x <0.5	0.5<= x <0.75	0.75<= x <1	1<=x

The module gives values for the Student management system project for WMC, DIT, CBO and RFC as 1.25, 2,

• 6.5,    and 1.3 respectively which is shown in following Fig 3. Similarly other modules are calculated.

From above snapshot we have values for WMC, CBO and DIT and RFC. Similarly other modules will be calculated in same manner.

• B.    Determining the Fuzzy Appraisal Matrix

From the Fig 2, i.e. snapshot of a one module, values for each metrics (Complexity, Coupling, Inheritance and Message Passing) can be calculated. Similarly for remaining 15 modules the method will be same. Further these values are compared from the Table 2 which has provided the ranges for four of metrics and then differentiated accordingly for 16 modules. Table 3 is showing the Membership Mapping for the metrics.

Table 3. The Membership Mapping for Task Ranking

COMP	Excellent	Good	Medium	Poor	Very Poor
M1					×
M2			×
M3				×
M4				×
M5			×
M6					×
M7			×
M8					×
M9			×
M10					×
M11	×
M12					×
M13					×
M14		×
M15				×
M16			×
Total	1	1	5	3	6
Ri	0.0625	0.0625	0.3125	01875	0.375

R -	COMP	0.063	0.063	0.313	0.1 ss	0.373
	COUP	025	0.123	0	0.123	0.25
	INH	0	0	0	0	0
	MSGP	1	0	0.63	0375	0

Fig. 3. Matrix representation for Final Appraisal(R)

Where COMP-Complexity, COUP-Coupling, INH-Inheritance, MSGP-Message Passing

Previously Defined, Weight by equation (i) gives

A= [.226   .262     .454     .058]

And so for B= [A] (eqn 1) * [R] (Table 3), the values are b1 b2 b3    b4 b5

B = [0.203 0.047 0.354 0.246 0.150]             (ii)

Where B is the Appraisal Vector . This is the final appraisal vector.

• C.    Defuzzification

The procedure of converting fuzzy numbers to their approximate crisp numbers is referred to as defuzzication . It is defuzzified to a comprehensive score. This paper has defined excellent, good, medium, poor, very poor in appraisal grading as 30, 45, 60, 75, and 90 respectively so appraisal vector B can be defuzzified according to the formula [37, 38]:

2    bi2 a.

M b = ^--

2    bi2

i = 1

After defuzzifying, M b = 60.226

• V. Conclusion and Future Scope

Presented paper proposed a model to evaluate the maintainability of object-oriented software system. The inputs for the method were complexity, coupling, inheritance and message passing, which affect the maintainability of the software in different object-oriented software system. These inputs were determined on the basis of survey from different experts project managers, systems developers, researchers and other who are working in this field. A maintainability model is proposed and the maintainability of object oriented system is evaluated based on this model. A fuzzy criteria and AHP technology and fuzzy methodology were used to evaluate the maintainability. The proposed model will be helpful to find the maintainability as a single score for the object oriented software system. Future study will focus on maintainability evaluation of some industrial software system.