Hindawi

Mathematical Problems in Engineering

Volume 2020, Article ID 3428604, 20 pages
https://doi.org/10.1155/2020/3428604

Research Article
Method of Coupling Metrics for Object-Oriented Software System
Based on CSBG Approach

Aihua Gu ,1 Lu Li,1 Shujun Li,1 Qifeng Xun,1 Jian Dong,1 and Jianhong Lin2
1
School of Information Engineering, Yancheng Teachers University, Yancheng 224002, China
2
Zhejiang Ponshine Information Technology Co., Ltd., Hangzhou 310012, China

Correspondence should be addressed to Aihua Gu; [email protected]

Received 2 November 2019; Accepted 5 February 2020; Published 19 March 2020

Guest Editor: Weifeng Pan

Copyright © 2020 Aihua Gu et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Context. Coupling between classes is an important metric for software complexity in software systems. Objective. In order to
overcome the shortcomings of the existing coupling methods and fully investigate the weighted coupling of classes in different
cases in large-scale software systems, this study analyzed the relationship between classes at package level, class level, and method
level. Method. The software system is considered as a set of special bipartite graphs in complex networks, and an effective method
for coupling measurement is proposed as well. Furthermore, this method is theoretically proved to satisfy the mathematical
properties of coupling measurement, leading to overcome the disadvantages of the majority of existing methods. In addition, it
was revealed that the proposed method was efficient according to the analyses of existing methods for coupling measurement.
Eventually, an algorithm was designed and a program was developed to calculate coupling between classes in three open-source
software systems. Results. The results indicated the scale-free characteristic of complex networks in the statistical data. Addi-
tionally, the calculated power-law value was used as a metric for coupling measurement, so as to calculate coupling of the three
open-source software. It indicated that coupling degrees of the open-source software systems contained a certain impact on
evaluation of software complexity. Conclusions. It indicated that coupling degrees of the open-source software systems contained a
certain impact on evaluation of software complexity. Moreover, statistical characteristics of some complex networks provided a
reliable reference for further in-depth study of coupling. The empirical evidence showed that within a certain range, reducing the
coupling was helpful to attenuate the complexity of the software, while excessively blindly pursuit of low coupling increases the
complexity of software systems.

1. Introduction focuses on the objects of a problem throughout develop-

ment. Interactions between objects reflect the interdepen-
Coupling refers to the degree of interdependence between dence between objects. If objects are isolated, then the
software modules; a measure of how closely connected two software system can only achieve simple functions. How-
routines or modules are [1]; and the strength of the rela- ever, objects are equivalent to cells in human body. If cells
tionships between modules. Structured design, including are completely isolated from human body, they basically do
cohesion and coupling, was published in an article by Ste- not play any significant role, reflecting that functions of a
vens et al. and a book by Stevens et al. [2, 3], and the latter software system require a tight coupling between objects. On
subsequently became standard terms. Coupling is consid- the other hand, tight coupling between objects would lead to
ered as a double-edged sword in object-oriented pro- a water-wave effect, meaning that changes in one object may
gramming. On the one hand, object-oriented software result in further changes in other objects. The most terrible
development (OOSD) includes object-oriented requirement case is that there is a possibility of “avalanche” effect, which
analysis, as well as object-oriented design. OOSD is a may affect the whole system, leading to a sharp decline in the
practical method of developing a software system which testability, understandability, reliability, and maintainability
2 Mathematical Problems in Engineering

of the system. Therefore, it is expected that classes are loosely methods contain some limitations, which cannot
coupled in terms of software design. A system can be tightly properly satisfy the requirement of an effective
coupled in one aspect while being loosely coupled in an- coupling measurement in complex software systems.
other. However, software developers mainly prefer to de- (3) Although a number of coupling measurement
velop those systems that are as loosely coupled as possible; methods analyze network relationship from overall
thus, design, testing, and maintenance of the system would and macro perspectives based on graph theory, the
be relatively independent and more reasonable. Moreover, a majority of measurement metrics mainly use classes,
decrease may be observed in the possibility that errors packages, or methods as nodes to construct some
propagate between modules if there are few connections undirected, directed, unweighted, or weighted net-
between modules [4]. Coupling has been widely used in work models. Moreover, they ignore a complex re-
evaluation of the degree of failure in classification [5–7], lationship of object-oriented characteristics between
effective analysis [8, 9], and design pattern [10] of software different classes. Some methods have not been
systems. theoretically verified for developing the mathemat-
The present article has the following organization: ical characteristics of the measurement metrics.
Section 2 summarizes the materials and methods. Section 3
shows the results. Section 4 provides a conclusion and The process of establishment of an effective method for
suggests perspectives. coupling measurement between classes in a software sys-
tem is determined by the following two aspects: reasonable
measurement metrics and theoretical support of mea-
2. Materials and Methods surement metrics [24, 25]. Briand et al. mathematically
2.1. Methods. Currently, the methods for coupling mea- analyzed measurement metrics of the software system and
surement of object-oriented software systems are mainly presented a robust theoretical support for the measurement
structure-oriented measurement methods (Tables 1 and 2) metrics [4, 26, 27]. Many of complex networks have been
[8, 11–15]. shown to share the features such as “scale-free” and “small
Comparative analysis of typical methods is shown in world” [28, 29]. Pan et al. revealed many physics-like laws
Tables 1–3, indicating that in software systems from a complex network perspective
recently [30, 31]. Studies on complex networks and soft-
(1) Methods for coupling measurement are mainly ware engineering revealed that class-level, method-level,
based on method invocation between classes. and package-level diagrams of a software system could
(2) Calculation of coupling strength is defined as the show the characteristics of “scale-free” and “small world,”
degree of method invocation, which is weighted which provided a novel perspective for finding more
coupling. reasonable measurement metrics [32–34]. Complex net-
(3) A small number of methods use fan-in and fan-out as work theories were applied to measure software [35, 36],
metrics. identify key software elements [37], and cluster Web ser-
vices [38–39]. Researchers have found that many real
(4) Inheritance is dominant. networks have the bipartite graph characteristics of com-
(5) Few methods use static method invocation, system plex networks [40–45]. With combination of package level,
measurement, and package-level metrics. class level, and method level, this study analyzed a complex
In addition to the abovementioned structural informa- relationship between classes in the same layer and all layers
tion methods, some scholars have recently used dynamic of a package and proposed a method for coupling metrics
information methods [17, 18], semantic information for object-oriented systems based on bipartite graph of
methods [19–21], and logical information methods [22, 23]. complex networks, named here CSBG, and object-oriented
Based on the results of previous studies, methods of coupling software systems were expressed as a set of special bipartite
measurement cover the following cases: graphs.

(1) The DCMs are more accurate than that of structural

information methods, while it seems to be difficult in 2.2. Problem Description. In this study, a statistical method
the measurement of coupling metrics. However, for complex networks was used to analyze the degree of fan-
structural information methods are more intuitive out and the heterogeneity of classes at the same layer and all
and easier to be perceived compared with semantic layers of a package, in addition to the calculation of coupling
information and logical information methods. of software systems.
(2) At present, the majority of the traditional structural
information methods analyze coupling based on the
degree of connecting edges between classes and 2.2.1. Relationship between Classes
mainly focus on complexity between classes and
emphasize more on measurement from a local fine- Definition 1. ASS relationship (association and aggregation).
grained aspect. Moreover, these degrees of con- Association means which/how classes interact with each
necting edges only consider a certain or a limited other, and association can be represented by a line between
aspects of software engineering. Therefore, these these classes with an arrow indicating the navigation
Mathematical Problems in Engineering 3

Table 1: The first part of existing methods for coupling measurement.

Method Description
CBO(c) � |{d ∈ C − {c} | uses(c, d)∨uses(c, d)}|; the metric is 1, if method in one class invokes other
CBO [11]
classes or is attributed to another class, otherwise it is 0
CBO′ (c) � |{d ∈ C − ({c∪Ancestors(C)}) | uses(c, d)∨uses(c, d)}|; this is similar to CBO method;
CBO′ [12]
however, that does not consider inheritance
RFC(c) � RFC1 (c), which is used for calculating the number of methods responding to an object’s
RFC [13]
message
R0 (c) � M(c), Ri+1 (c) � ∪m∈Ri (c)PIM(m), that is, a set of polymorphic methods invoked by functions in
RFCα [13]
set Ri (c); then RFCα (c) � |∪αi�0 Ri (c)| with α � 1, 2, 3, . . . ,
RFC′ [13] RFC′ (c) � RFC∞ (c)
MPC(c) � 􏽐m∈MI (c)􏽐m′ ∈SIM(m)− MI (c)NSI(m, m′ ); this calculates the number of static method invocation of
MPC [13]
classes
DAC [14] DAC(c) � |􏼈a | a ∈ AI (c)∨T(a) ∈ C􏼉|
DAC′ (c) � |􏼈T(a) | a ∈ AI (c)∨T(a) ∈ C􏼉|; this formula is similar to DAC; however, if there is a
DAC′ [14]
relationship between classes, the metric is 1, otherwise the metric is 0

Table 2: The second part of existing methods for coupling measurement.

Method Description
COF [14] COF(C) � (􏽐c∈C |{d | d ∈ C − {c}∪Ancestors(c)∧uses(c, d)}|)/|C|2 − |C| − 2􏽐c∈C Descendent(c)
ICP [14] This method calculates the parameters invoked by the method in a weighted class
IH-ICP [14] This is similar to ICP, however, that only considers inheritance
NIH-ICP [14] This is similar to ICP, however, that does not consider inheritance
SIMAS [8] This method calculates the number of direct or indirect invocations between static methods of two different classes
PIM [8] This method calculates the number of invocations in class C of methods in class D, and polymorphism is considered
This method calculates the number of direct or indirect invocations between class methods, and polymorphism is taken
PIMAS [8]
into account
INAG [8] The metric is 1 if there is an indirect aggregation between two classes; otherwise, the metric is 0
ACAIC(c) � 􏽐d∈Ancestors(c) CA(c, d); this calculates the number of out-degrees between one class and attributes of
ACAIC [15]
another classes in two classes with inheritance
OCAIC(c) � 􏽐d∈Others(c)∪Friends(c) CA(c, d); it calculates the number of out-degrees between one class and attributes of
OCAIC [15]
another class in two classes without inheritance
ACMIC(c) � 􏽐d∈Ancestors(c) CA(c, d); it calculates the number of out-degrees between one class and methods of another
ACMIC [15]
class in two classes with inheritance
OCMIC(c) � 􏽐d∈Others(c)∪Friends(c) CA(c, d); it calculates the number of out-degrees between one class and methods of
OCMIC [15]
another class in two classes without inheritance
AMMIC(c) � 􏽐d∈Ancestors(c) MM(c, d); it calculates the number of out-degrees for methods between two classes with
AMMIC [15]
inheritance
OMMIC [15] OMMIC(c) � 􏽐d∈Others(c)∪Friends(c) MM(c, d)
ICF, FCF [16] ICFi � 􏽐nk�1 I(k, i)ICFk , I(i, j) � e(i, j)/ 􏽐nk�1 e(i, k), FCFi � 􏽐nk�1 F(k, i)FCFk , I(i, j) � e(i, j)/ 􏽐nk�1 e(k, j)

direction. Aggregation implies that one class exists in an- subpackages. Software systems can actually be considered to
other class in the form of attribute. be a tree hierarchical structure composed of packages.
t layer of a package is defined as pt ≤ Et+1 , Rt+1 > . Et+1
represents a set of classes in the t layer, while this layer does
Definition 2. DEP relationship (dependency)
not contain subpackages of this layer. Rt+1 represents class
DEP_D: dynamic dependency refers to an instance method
relationship in the t layer, that is, Rt+1 ⊆ Et+1 × Et+1 .
in a class that invokes methods and attributes in another class.
S layer(i) is defined as a set of weighted fan-out of Ci at
DEP_S: static dependency refers to static methods in a
the t layer of the package, that is, S layer(i) ⊆ Rt+1 . S all(i) is
class invokes methods and attributes in another class.
the set of weighted cross-package fan-out of Ci , that is,
S all(i) ⊆ R1 ∪ R2 · · · Rt · · ·.
Definition 3. GEN (generalization)
One class inherits with another class, or one class imple-
ments interfaces with another class, or that of an abstract class. 2.3. CSBG for Coupling Measurement. Software stability and
modularity could be measured based on complex network
theories. In this study, software systems can be expressed as a
2.2.2. Definition of Package Hierarchy. Packages of an ob- set of bipartite graphs that use nodes as classes, and ASS as
ject-oriented software system include classes and sub- well as DEP are the edges constituted by attributes of the
packages in the current hierarchy, and these subpackages class with those of another class based on complex network
contain classes in the current hierarchy and their theories. However, GEN is a direct connection between
 4

Table 3: Comparative analysis for the typical methods of coupling measurement.

Fan-out/ Indirect Static System Package
Method Type Strength Inheritance Weighted
fan-in coupling invocation metric level
CBO Method Both No Both No No
invocation, #coupled classed Non-inh.-
CBO′ No No
attribute reference based
RFC No Yes
RFCα Method invocation #methods invoked Import Yes Both Yes No No
RFC′ Yes Yes
MPC Method invocation #methods invocations Import No Both Yes Yes
DAC #attributes Yes No
Type of attribute Import No Both
DAC′ #distinct types No No
Method
Non-inh.-
COF invocation, #coupled classed Both No Yes No Yes
based
attribute reference
ICP Both Yes No
#methods
IH-ICP inh.-based Yes No
Method invocation invocations, Import No
Non-inh.- No
NIH-ICP #parameters passed Yes No
based
SIMAS Method invocation #methods invocations Yes Both Yes Yes
PIM Method invocation #methods invocations No Both Yes No
Import
PIMAS Method invocation #methods invocations Yes Both Yes No
INAG Type of attribute #attributes Yes Both No No
ACAIC inh.-based Yes No
Type of attribute #attributes Non-inh.-
OCAIC Yes
based
ACMIC inh.-based Yes
Type of parameter #of parameters Import No Non-inh.- No
OCMIC Yes
based
AMMC inh.-based Yes
Method invocation #method invocations Non-inh.-
OMMC Yes
based
ICF Method Import Yes
invocation, #method invocations No Both No No No
FCF Export Yes
attribute reference
Mathematical Problems in Engineering
Mathematical Problems in Engineering 5

the same layer of the package (see step 3,

(|􏽐S layer(i)|)), and the weighted out-degree
((| 􏽐 S all(i)|)) of all classes with classes across layers
of the package (see step 3).
(3) ASS layer, DEP D layer, DEP S layer, and
GEN layer at the same layer of the package were
calculated. S layer is determined by adding the
weights of x1 , x2 , x3 , and x4 to the four mentioned
metrics, respectively, in order to calculate ASS all,
DEP D all, DEP S all, and GEN all between classes
across different layers of the package. Then, weights
of x5 , x6 , x7 , and x8 were added to these four metrics
to determine S all:
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
􏼌􏼌 􏼌 􏼌􏼌 􏼌􏼌 􏼌
􏼌􏼌􏽘 S layer(i)􏼌􏼌􏼌 � ⎡ ⎣ 􏼌􏼌􏽘 ASS layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP D layer(i)􏼌􏼌􏼌,
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌
􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
􏼌􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌
􏼌􏼌􏽘 DEP S layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 GEN layer(i)􏼌􏼌􏼌⎦⎤
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌
i�1 i�1

× 􏼂x 1 , x2 , x3 , x4 􏼃 T ,
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
Figure 1: Illustration of a software system network composed of a 􏼌􏼌 􏼌 􏼌􏼌 􏼌􏼌 􏼌
􏼌􏼌􏽘 S all(i)􏼌􏼌􏼌 � ⎡
⎣ 􏼌􏼌􏽘 ASS all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP D all(i)􏼌􏼌􏼌,
set of special bipartite graphs (the large squares represent packages, 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌
and the 4 packages are at the same layer. The small squares show 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
subpackages in the package. The circles denote classes in the 􏼌􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌
package, and edges represent relationship between classes). 􏼌􏼌􏽘 DEP S all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 GEN all(i)􏼌􏼌􏼌⎤⎦
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌
i�1 i�1

classes. Therefore, object-oriented software systems are taken × 􏼂 x5 , x6 , x7 , x8 􏼃 T .

into account as a set of special bipartite graphs constituted by (1)
classes in the package, as shown in Figure 1. There are defects
in the coupling metrics containing the two metrics of fan-in (4) S layer and S all are weighted to calculate the
and fan-out, because their total number is equal in a software weighted out-degree S of the software system. The
system [46]. Therefore, this study only analyzed fan-out system coupling is calculated through dividing S by
metric. The coupling strength between classes is correlated the number of classes:
with the complexity of information exchange between mod- 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌
ules. The more complex the information interaction (such as S � 􏼌􏼌􏼌􏽘 S(i)􏼌􏼌􏼌 � 􏼌􏼌􏼌􏽘 S layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 S all(i)􏼌􏼌􏼌􏼌⎤⎦ ×[α, β]T ,
􏼌 􏼌 ⎣
⎡ 􏼌 􏼌 􏼌
CBO), the tighter the coupling [47]. Coupling measurement 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌
metrics refer to the weighted fan-out of classes in special
bipartite graphs. In these special bipartite graphs, classes as- S
S� .
sociate with a class that may be at the same layer of the same n
package or at different layers of the package. Therefore, this (2)
study analyzed degrees of fan-out for classes in the same layer
and all layers of the package. Moreover, heterogeneity of the
abovementioned weighted out-degree was analyzed.
2.3.2. Calculation of the Weighted Fan-Out of Classes in a
Software System. A special bipartite graph is constructed
2.3.1. Demonstration of CSBG for Coupling Measurement. between classes of a software system. Weighted fan-out of
The detailed scheme proposed here is explained in the classes in the special bipartite graph is analyzed based on
following, as illustrated in Figure 2: characteristics of the object-oriented software.
(1) The object-orient software systems are constructed (1) Construction of the Special Bipartite Graph in Software
as directed weighted network graphs, and classes and Systems. In the graph G(V, E), if we divide the set V of nodes
relationship between classes are shown as nodes and into two complementary subsets S and T,
edges, respectively. S∪T � V, and S ∩ T � ϕ, the graph G(V, E) is the bipartite
(2) The package level, class level, and method level are graph. In the graph Gij (Cij , Eij ) constructed by classes Ci
combined to construct special weighted bipartite and Cj for the software, if only coupling relationship be-
graphs between classes, aiming to make preparation tween classes is considered, coupling of methods and at-
for calculating the weighted out-degree of classes at tributes in the class wouldn’t be taken into account; then, the
6 Mathematical Problems in Engineering

–
S

1/n

∑S(i)
β, b
b4
b3+ 5 +b
b2+ 6 +b
α, b1+ 7 +b
8

∑S_layer(i) ∑S_all(i)

x1, b1 x4, b4 x5, b5 x8, b8

x6, b6 x7, b7
x2, b2 x3, b3

∑ASS_layer(i) ∑DEP_D_layer(i) ∑DEP_S_layer(i) ∑GEN_layer(i) ∑ASS_all(i) ∑DEP_D_all(i) ∑DEP_S_all(i) ∑GEN_all(i)

Figure 2: Illustration of the coupling measurement for a software system.

property of bipartite graph Ci ∩ Cj � ∅ is satisfied. A net- ASS DEP_D DEP_S GEN

work diagram constructed by classes Ci and Cj satisfies the
following formula: Ci ∪Cj � Cij . Moreover, the two points of
(Ci, Cj) Gij (Cij, Eij) Ci ∩ Cj = Ø
a connecting edge between classes Ci and Cj are in classes Ci
and Cj , respectively.
In summary, the complex coupling between classes Ci and Ci ∪ Cj = Cij
Cj constructs a bipartite graph Gij (Cij , Eij ). However,
Gij (Cij , Eij ) not only possesses the general properties of a G = (C1, C2, …, CN) Rt+1 ⊆ Et+1 × Et+1
bipartite graph, including method invocation and depen-
dencies, but also possesses its own characteristics. In aggre-
gation, reference, inheritance, and interface implementation
between classes Ci and Cj , the two points of the connecting
S_all(i) ⊆ R1 ∪ R2… Rt… S_layer(i) ⊆ Rt+1
edge are in classes Ci and Cj , respectively. This bipartite graph
Gij (Cij , Eij ) is defined as a special bipartite graph. However, Figure 3: Abstract diagram of the software system network
the software system G(C, E) can be considered as a set of composed of a set of special bipartite graphs.
special bipartite graphs Gij (Cij , Eij ) as well (Figure 3).
In the present study, the coupling of the complete bipartite
graph Gij (Cij , Eij ) constructed by classes Ci and Cj was used In the class Ci , instantiated object Ojp′ of class Ci was
to analyze the coupling of the software system G(C, E). implemented (1 ≤ p′ ≤ p), or Ojp′ was a part of class
Ci ; then, there was an ASS edge between classes Ci and
(2) Modeling the Coupling of Special Bipartite and Calcu- Cj , that is, (Ci , Cj Ojp′ ) ∈ RASS . The set of ASS
lating the Number of Weighted Fan-Out in Software Systems. weighted fan-out of class Ci was
In this study, a software system G � (C1 , C2 , . . . , CN ) was 􏼌􏼌
defined. Classes Ci and Cj were defined as two different ASS(i) � 􏼚􏼐Ci , Cj Ojp′ 􏼑 􏼌􏼌􏼌 1 ≤ j ≤ N, 1 ≤ p′ ≤ p,
classes in a software system: Ci � (Oi , Ai , Mi ). Among them, (3)
Oi � 􏽮Oi1 , Oi2 , . . . , Oip 􏽯 was the set of instantiated objects in 􏼐Ci , Cj Ojp′ 􏼑 ∈ RASS 􏼛.
class Ci and p is the number of instantiated objects. Ai �
􏽮Ai1 , Ai2 , . . . , Aiq 􏽯 is the attribute set of class Ci , and q is the DEP
number of attribute. Mi � 􏼈Mi1 , Mi2 , . . . , Mir 􏼉 is the Relationship between classes is implemented defined
method set of class Ci , and r is the number of methods. The by instance methods and variables.
methods included class methods and instance methods, that In class Ci , there are instance methods of class Cj : if
is, (C M ∪ C O M) ⊂ M. Cj Ojp′ Mjr′ , 1 ≤ j ≤ N, 1 ≤ p′ ≤ p, and 1 ≤ r′ ≤ r,
The relationship of the special bipartite graph between then the relationship between classes Ci and Cj is
classes Ci and Cj can be summarized as follows: defined as (Ci , Cj Ojp′ Mjr′ ) ∈ RDEP D M . In class Ci ,
there are instance variables of class Cj : if
ASS
Cj Ojp′ Ajq′ , 1 ≤ j ≤ N, 1 ≤ p′ ≤ p, and 1 ≤ q′ ≤ q,
In class Ci , there was instantiation of class Cj (as- then relationship between classes Ci and Cj is defined
sociation), or one class existed in another class in the as (Ci , Cj Ojp′ Ajq′ ) ∈ RDEP D A . Under the condi-
form of attribute (aggregation), which was defined as tion of instance methods and instance variables, the
Cj Ojp′ , where 1 ≤ p′ ≤ p. set of weighted fan-out for Ci was
Mathematical Problems in Engineering 7

DEP D(i) � 􏽮 􏼐Ci , Cj Ojp′ Mjr′ 􏼑, 􏼐Ci , Cj Ojp′ Ajq′ 􏼑 number of weighted fan-out in a class. The third case is that
􏼌􏼌 the number of weighted fan-out of a class is heterogeneity,
· 􏼌􏼌 1 ≤ j ≤ N, 1 ≤ p′ ≤ p, 1 ≤ r′ ≤ r, which approximately obeys the power-law distribution. For
the second case, heterogeneity of the out-degree of the class
1 ≤ q′ ≤ q, 􏼐Ci , Cj Ojp′ Mjr′ 􏼑 ∈ RDEP D ,
is superior than that of the first case; however, this is im-
􏼐Ci , Cj Ojp′ Ajq′ 􏼑 ∈ RDEP D A 􏽯.
possible to be compared with the third case. For the third
case, because the number of fan-out is limited for the
(4)
majority of classes, only few classes have a large number of
fan-out; therefore, maintenance staff can dedicate more
Implementing connecting edges between two classes effort on these few classes. Moreover, the maintenance
through class methods and class variables. workload of these classes is lower than that of the first case.
If there are class methods of Cj (static methods) in class In this study, heterogeneity under the situation of fan-
Ci : Cj .Mjr′ , and 1 ≤ r′ ≤ r, then the relationship be- out was analyzed. If the distribution was the above-
tween classes Ci and Cj is defined as mentioned third case, then the larger the power-law value,
(Ci , Cj Mjr′ ) ∈ RDEP S M , 1 ≤ r′ ≤ r. If there are class the easier the maintenance, and the smaller the coupling
variables (static variables) of Cj in class degree. However, if the distribution was one of the other two
Ci :Cj .Ajq′ , 1 ≤ q′ ≤ q, then the relationship between cases, then it was considered in this study that the power-law
classes Ci and Cj is defined as value was equal to 1. b1 , b2 , b3 , and b4 are the power-law
(Ci , Cj Ajq′ ) ∈ RDEP S A , 1 ≤ q′ ≤ q. Thereafter, under values for the distribution of ASS layer, DEP D layer,
the conditions of class methods and class variables, the DEP S layer, and GEN layer, respectively. b5 , b6 , b7 , and b8
set of weighted out-degree for class Ci was are the power-law values for the distribution of ASS all,
DEP D all, DEP S all, and GEN all, respectively. The cal-
DEP S(i) � 􏽮􏼐Ci , Cj Mjr′ 􏼑, 􏼐Ci , Cj Ajq′ 􏼑 culating formula for weights was as follows:
􏼌􏼌
· 􏼌􏼌 1 ≤ j ≤ N, 1 ≤ r′ ≤ r, 1 ≤ q′ ≤ q,
T
(5) T
4
bi 4 bi 4 bi 4 bi ⎤⎦
􏼐Ci , Cj Mjr′ 􏼑 ∈ RDEP S M, 􏼂x1 , x2 , x3 , x4 􏼃 � ⎡⎣􏽘 ,􏽘 ,􏽘 ,􏽘 ,
b b b b
i�1 1 i�1 2 i�1 3 i�1 4
􏼐Ci , Cj Ajq′ 􏼑 ∈ RDEP S A 􏽯.
8 T
T b 8 b 8 b 8 b
GEN 􏼂x5 , x6 , x7 , x8 􏼃 � ⎡⎣􏽘 i , 􏽘 i , 􏽘 i , 􏽘 i ⎤⎦ ,
b b b b
i�5 5 i�5 6 i�5 7 i�5 8 (7)
As the inheritance is preferred in software engi-
neering, if one class is a subclass of another class, the 􏽐8i�4 bi
derived connecting edge was taken into account only α� ,
􏽐8i�1 bi
once in this study. Because transfer of derived con-
necting edges would make the software system net-
β � 1 − α.
work more complex, this study did not consider
transfer of derived connecting edge but only con- In the present study, statistical analyses were performed
sidered the conditions that class Ci was a direct for the out-degree of the three open-source software sys-
subclass of class Cj (through extension), or class Ci tems, and the distributions were the first and the third cases
was implemented through interface class Cj (through as mentioned above, demonstrating that the proposed
implementation), or class Ci was implemented by method had a certain practical value.
abstract class Cj (through extension). Thus, there was
a GEN connecting edge between classes Ci and Cj ,
which was defined as (Ci , Cj ) ∈ RGEN , and a set of 2.4. Theoretical Verification of Coupling Metrics. Whether the
GEN weighted out-degree for class Ci was proposed CSBG method met the mathematical properties of
the measurement metrics [4] was theoretically verified.
􏼌􏼌
GEN(i) � 􏼚􏼐Ci , Cj 􏼑 􏼌􏼌􏼌 􏼐Ci , Cj 􏼑 ∈ RGEN 􏼛. (6) CSBG Property 1. CSBG satisfies nonnegativity.

Proof. In an object-oriented software system G �

(c1 , c2 , . . . , cN ), there are two classes c1 , c2 ∈ G. When
(3) Determination of Weights. In software systems, one class ASS layerDEP D layer, DEP S layer, and GEN layer are all
can construct one or more special bipartite graphs with other 0, the minimum value of the software system CSBG(G) is 0.
classes. Supposing that the number of classes and the total However, there is a maximum value M(M > 0), so that the
number of weighted fan-out of all classes are definite in a CSBG(G) value is in the range of [0, M]. Thus, nonnegativity
software system, the first case is that the number of weighted of CSBG is satisfied, and the proposition is proved. □
fan-out in each class is the same or roughly the same. The
second case is that there is no rule for the distribution of the CSBG Property 2. CSBG satisfies zero value.
8 Mathematical Problems in Engineering

Proof. As described in CSBG property 1, if the minimum Graph1

value is 0, then CSBG satisfies zero-value, and the propo-
sition is proved as well. □ Graph.polygon2

CSBG Property 3. CSBG satisfies monotonicity. Triangle2 Shape3

Proof. If one edge is arbitrarily added in the system, the

weighted out-degree of classes would increase according to
CSBG measurement metrics. Obviously, the coupling in- Quadrilateral2 Point3
creases as well. Thus, CSBG meets monotonicity and the
proposition is proved. □
CSBG Property 4. CSBG meets the property of class Line3
Square2
merging.

Proof. In an object-oriented software system G �

(c1 , c2 , . . . , cN ), there are two classes c1 , c2 ∈ G, and class c′ is Figure 4: Diagram at package level.
a merger of classes c1 and c2 . The object-oriented system G′
is a system in which classes c1 and c2 in G are replaced by
class c′ . CSBG mainly calculates the weighted out-degree of
classes in software systems. Therefore,
􏼌􏼌
􏽨CSBG c1 􏼁 + CSBG c2 􏼁 ≥ CSBG c′ 􏼁 􏼌􏼌 CSBG(G) ≥ CSBG G′ 􏼁􏽩.
(8)
□
CSBG Property 5. CSBG satisfies the merge property of
two irrelevant classes.

Proof. In an object-oriented software system G �

(c1 , c2 , . . . , cN ), there are two classes c1 , c2 ∈ G, and the two
classes are not coupled. Moreover, class c′ is the merger of Figure 5: Demonstration of class shape.
classes c1 and c2 . The object-oriented system G′ is a system in
which classes c1 and c2 in G are replaced by class c′ . CSBG
mainly calculates the weighted out-degree of classes in
software systems. Therefore,

􏼂CSBG c1 􏼁 + CSBG c2 􏼁 � CSBG c′ 􏼁 | CSBG(G) � CSBG G′ 􏼁􏼃.

(9)
□

2.5. Comparative Experiment. In the next sections, the CSBG

method is herein proposed for coupling measurement and
the existing measurement methods were compared and
analyzed in order to verify the rationality of the results of
CSBG measurement.
Figure 6: Demonstration of class point.
2.5.1. Calculating the Coupling of the Software System Using
CSBG. In this section, CSBG for coupling measurement was
compared with the existing measurement methods. Among them, the first three classes were in package graph, and
This experiment was conducted on a simple system as an the last three classes were in package graph.polygon (hierarchy
example to analyze and compare the measurement values by the of classes in package level is shown in Figure 4). There were
existing coupling measurements. This system was composed of inheritance, combination, variable declaration, and method
6 classes (Shape.java, Point.java, Line.java, Triangle.java, invocation among these classes, which were appropriate for
Quadrilateral.java, and Square.java), which described shapes, analyzing the coupling model. Codes of classes are shown in
points, edges, triangles, quadrilaterals, and squares, respectively. Figures 5–10.
Mathematical Problems in Engineering 9

Figure 7: Demonstration of class line.

Figure 8: Demonstration of class triangle.

Figure 9: Demonstration of class square.

There were three classes in the package graph, including DEP_S, and GEN, were corresponded to the cases described
class Shape, class Point, and class Line. in Section 2.3. Out-degrees of classes in various layers are
There were three classes in the package graph.polygon, shown in Table 4.
which were classes of Triangle, Square, and Quadrilateral. The mathematical model described in Section 2.3 was
In this study, an algorithm was designed and the pro- herein used. Because the number of classes was small, the
gram was developed based on the aforementioned mathe- heterogeneity of out-degree of classes could not be reflected.
matical model, mainly calculating the four metrics for the Moreover, heterogeneity had little impact on the coupling in
out-degree of classes in the same layer and different layers of this example. Therefore, it was considered that heterogeneity
the package in software systems, including ASS, DEP_D, was approximately the same. Coupling of software systems
DEP_S, and GEN. Coupling metrics, including ASS, DEP_D, was calculated as follows:
10 Mathematical Problems in Engineering

Figure 10: Demonstration of class quadrilateral.

􏽐6i CSBG Ci 􏼁
CSBG(G) � � 6.55,
6
T
􏼂x1 , x2 , x3 , x4 􏼃 � [1, 1, 1, 1]T ,
T
􏼂x5 , x6 , x7 , x8 􏼃 � [1, 1, 1, 1]T ,

α � 0.5,

β � 1 − α � 0.5,
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
􏼌􏼌 􏼌 􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌
􏼌􏼌􏽘 S layer(i)􏼌􏼌􏼌 � ⎡⎣􏼌􏼌􏼌􏽘 ASS layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP D layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP S layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 GEN layer(i)􏼌􏼌􏼌⎤⎦ × 􏼂x , x , x , x 􏼃T
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 1 2 3 4
􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌

� [6, 10, 0, 3] ×[1, 1, 1, 1]T

� 19,
(10)
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
􏼌􏼌 􏼌 􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌
􏼌􏼌􏽘 S all(i)􏼌􏼌􏼌 � ⎡⎣􏼌􏼌􏼌􏽘 ASS all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP D all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP S all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 GEN all(i)􏼌􏼌􏼌⎤⎦ × 􏼂x , x , x , x 􏼃T
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 5 6 7 8
􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌

� [24, 38, 0, 5] ×[1, 1, 1, 1]T

� 67,
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌
S � 􏼌􏼌􏼌􏽘 S(i)􏼌􏼌􏼌 � ⎡⎣􏼌􏼌􏼌􏽘 S layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 S all(i)􏼌􏼌􏼌􏼌⎤⎦ ×[α, β]T
􏼌 􏼌 􏼌 􏼌 􏼌
􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌

� [19, 67] ×[0.5, 0.5]T

� 43,

S 43
S� � � 7.17.
n 6
Mathematical Problems in Engineering 11

Table 4: Out-degrees of classes at the same layer and all layers of the package.
Out-degree at the same layer Out-degree of all layers
Class name
ASS DEP_D DEP_S GEN Total ASS DEP_D DEP_S GEN Total
Quadrilateral 0 0 0 0 0 8 16 0 1 25
Triangle 0 0 0 0 0 6 12 0 1 19
Line 6 10 0 1 17 6 10 0 1 17
Point 0 0 0 1 1 0 0 0 1 1
Shape 0 0 0 0 0 0 0 0 0 0
Square 0 0 0 1 1 4 0 0 1 5
Total 6 10 0 3 19 24 38 0 5 67

2.5.2. Analysis of the Results of Various Methods for Coupling and (4) the source codes were available as well. Scholars can
Measurement. Coupling of software systems was calculated freely download the source codes from an open-source
based on existing measurement methods, which is shown in website (http://sourceforge.net).
Table 5. In addition to CSBG coupling measurement, other
measurement methods mainly focus on measurement of a
certain local fine-grained aspect. These methods were based on 3.2. Association of Coupling with Statistical Characteristics of
the theory of reductionism, which did not investigate the the Three Open-Source Software Systems. Firstly, the pro-
coupling of software systems from an overall and global gram was developed and out-degree of classes at the same
perspective. Therefore, the measurement values were mostly layer and all layers of the package was eventually obtained,
either very large or very small, and several metrics were equal to including ASS, DEP_D, DEP_S, and GEN.
0. In addition, discrimination of these metrics was not sig- In this section, DEP_D and DEP_S were analyzed, and
nificant compared with other methods for coupling mea- the results are shown in Figures 11–18. In the experimental
surement. The metrics calculated by CSBG had a better results, DEP(i) was a nonstandardized part of probability
discrimination. Therefore, the existed methods have limita- distribution P(i). If, P(i) ∼ i− c , then DEP ∼ (i)− c . A linear
tions, which cannot accordingly satisfy an effective coupling function was fitted using the double logarithmic method that
measurement for complex software systems. CSBG not only was fitted to estimate Gamma index c (R is the Pearson’s
can consider a complex relationship between classes in object- correlation coefficient and SD is standard deviation; c is also
oriented software systems but also analyze the complexity of expressed as B in the following table).
classes and the special bipartite graph composed of classes from Although inheritance between classes increases coupling of
the prospective of overall package level. Therefore, the CSBG the system, this is encouraged by the software system, which is
measurement method contained a certain rationality in theory. conducive to reduce function definition and attribute defini-
tion in order to create a new class; thus, it is a poor coupling. It
3. Results can be seen from linear distribution of GEN (Table 6) that
neither all classes have an inheritance relationship, nor the
3.1. Application of CSBG Measurement Metrics in the Three GEN fan-out of all classes were very large or very small.
Open-Source Software Systems. In order to further validate However, classes with values equal to 0 or 1 were dominant.
the effects of CSBG, this study used CSBG to measure and Pearson’s correlation coefficient (R) and SD value pro-
analyze coupling between classes in the three Java open- vided the quality of the linear fitting; the larger the R value,
source software systems from different fields, including Art the better the quality of the linear fitting, and B is estimated
of Illusion [48], JabRef [49], and GanttProject [50]. Some Gamma index c. Moreover, the smaller the SD value, the
studies have reported results of class cohesion metrics for the better the quality of the linear fitting. As shown in Table 7, if
three open-source software systems [51–53]; it is feasible to 0.95 is considered to be the minimum value, it can be ap-
further study the complexity of the three open-source proximated that the distribution obeyed the power-law
software systems if there is a more reasonable method for distribution except that ASS value in JabRef was relatively
coupling measurement. In order to verify the effects of the small (0.91651 and 0.88148). Furthermore, the distributions
CSBG measurement method in actual open-source software of ASS layer, ASS_all layer, DEP_D layer, DEP_S layer,
systems, three Java open-source software systems from DEP_D_all layer, and DEP_S_all layer were assumed to obey
different aforementioned fields were used. Art of Illusion is a power-law distribution. The results demonstrated that there
software system for 3D rendering, modeling, and animation. was a certain rule for the number of fan-out of classes in the
JabRef is a graphical application for managing bibliographic form of ASS and DEP, which was not the case that the values
database. GanttProject is a software system for project were mostly large or small. However, they had “scale-free”
scheduling characterized by resource calendar, manage- property for complex networks, which obeyed the power-
ment, and import or export (MS Project, PDF, HTML). The law distribution. In actual software development process, if
reasons to use the three open-source software systems in the software developers excessively pursue low coupling be-
measurement were because (1) these systems were based on tween classes, a class may be divided into two or more
object-oriented Java; (2) the classes in the systems had a subclasses; thus, system complexity may be accordingly in-
certain scale; (3) the three systems were from different fields; creased. The process of determination of the range of coupling
12 Mathematical Problems in Engineering

Table 5: Results of various methods for coupling measurement.

Quadrilateral Triangle Line Point Shape Square Software system
CSBG 17.5 13.3 5.1 0.3 0 3.1 7.17
CBO 3 2 2 5 4 2
CBO′ 1 1 1 4 0 1
RFC 0 0 0 38 0 0
RFCα 0 0 0 38 0 0
RFC′ 0 0 0 38 0 0
MPC 0 0 0 0 0 0
DAC 8 6 6 0 0 4
DAC′ 1 1 1 0 0 1
COF 0.2
ICP 16 12 10 0 0 1
IH-ICP 0 0 0 0 0 1
NIH-ICP 16 12 10 0 0 0
SIMAS 0 0 0 0 0 0
PIM 16 12 10 0 0 1
PIMAS 16 12 10 0 0 1
INAG 1 1 1 0 0 1
ACAIC 0 0 0 0 0 0
OCAIC 4 3 3 0 0 0
ACMIC 0 0 0 0 0 0
OCMIC 0 0 0 0 0 0
AMMC 0 0 0 0 0 1
OMMC 16 12 10 0 0 1
ICF 0 0 0 1 0 0
FCF 1 1 1 0 0 1

y = 2.51614 ∗ x – 5.48165 100

y = 7.1483 ∗ x – 19.22832 10 y = 7.22368 ∗ x – 20.54354
R = 0.97722
log10 (ASS_layer(i))

R = 0.94567
log10 (ASS_layer(i))

R = 0.91651
log10 (ASS_layer(i))

10 SD = 0.08465
SD = 0.17962 SD = 0.11961
1 < i < 471 10
0 < i < 695 0 < i < 947
1
1
1
0.1
0.1
0.1 10 100 1000
1 10 100 1 10 100
log10 (i) log10 (i)
log10 (i)
(a) (b) (c)

Figure 11: The double logarithmic diagrams of the fan-out of ASS invocation for classes at the same layer of a package.

y = 6.07342 ∗ x – 14.40843 y = 20.06268 ∗ x – 55.32035 100 y = 25.12239 ∗ x – 73.32253

log10 (DEP_D_layer(i))

log10 (DEP_D_layer(i))

100 R = 0.97537
log10 (DEP_D_layer(i))

100 R = 0.97249 R = 0.95689

SD = 0.13761 SD = 0.12677 SD = 0.14931
0 < i < 471 0 < i < 695 10 0 < i < 947
10
10

1
1
1

1 10 100 1 10 100 10 100 1000

log10 (i) log10 (i) log10 (i)

(a) (b) (c)

Figure 12: The double logarithmic diagrams of the fan-out of DEP_D invocation for classes at the same layer of a package.
Mathematical Problems in Engineering 13

y = 5.52632 ∗ x – 13.42274 y = 14.97997 ∗ x – 41.00201 10 y = 17.20075 ∗ x – 50.19695

log10 (DEP_S_layer(i))
log10 (DEP_S_layer(i))
log10 (DEP_S_layer(i))

R = 0.95388 100 R = 0.94658 R = 0.94086

10 SD = 0.14533 SD = 0.19504 SD = 0.1309
0 < i < 471 0 < i < 695 0 < i < 947
10 1
1

1 0.1
0.1
1 10 100 1 10 100 10 100 1000
log10 (i) log10 (i) log10 (i)

(a) (b) (c)

Figure 13: The double logarithmic diagrams of the fan-out of DEP_S invocation for classes at the same layer of a package.

3.0 4 3.0
2.5 2.5
3
GEN_layer(i)

GEN_layer(i)

GEN_layer(i)
2.0 2.0
1.5 2 1.5
1.0 1.0
1
0.5 0.5
0.0 0 0.0
0 100 200 300 400 500 –100 0 100 200 300 400 500 600 700 0 200 400 600 800 1000
i i i
(a) (b) (c)

Figure 14: The double logarithmic diagrams of fan-out of GEN invocation for classes at the same layer of a package.

100 y = 1.84955 ∗ x – 3.37653 100 y = 1.65033 ∗ x – 3.69578 y = 3.64204 ∗ x – 9.71088

R = 0.97418 R = 0.88148 R = 0.94905
log10 (ASS_all(i))
log10 (ASS_all(i))

log10 (ASS_all(i))

SD = 0.10346 SD = 0.2079 10 SD = 0.13485

10 0 < i < 471 10 0 < i < 695 0 < i < 947

1
1 1

0.1
1 10 100 1 10 100 10 100 1000
log10 (i) log10 (i) log10 (i)

(a) (b) (c)

Figure 15: The double logarithmic diagrams of the fan-out of ASS invocation for classes at all layers of a package.

y = 3.39056 ∗ x – 7.06826 y = 8.47787 ∗ x – 22.41636 y = 9.56684 ∗ x – 26.95122

100 R = 0.97314 100
log10 (DEP_D_all(i))

R = 0.97028 R = 0.96602
log10 (DEP_D_all(i))

log10 (DEP_D_all(i))

100 SD = 0.14556 SD = 0.13056 SD = 0.13634

0 < i < 471 0 < i < 695 0 < i < 947
10 10
10
1 1
1
1 10 100 1 10 100 10 100 1000
log10 (i) log10 (i) log10 (i)

(a) (b) (c)

Figure 16: The double logarithmic graph of fan-out of DEP_D invocation for classes at all layers of a package.
14 Mathematical Problems in Engineering

100 y = 3.5066 ∗ x – 7.66929 y = 4.47638 ∗ x – 11.10457 y = 5.99337 ∗ x – 16.67654

100

log10 (DEP_S_all(i))
log10 (DEP_S_all(i))

log10 (DEP_S_all(i))
R = 0.97265 R = 0.96831 R = 0.95419
SD = 0.1208 SD = 0.14207 10 SD = 0.12876
10 0 < c < 471 0 < i < 695 0 < c < 947
10
1
1
1
0.1
1 10 100 1 10 100 10 100 1000
log10 (i) log10 (i) log10 (i)

(a) (b) (c)

Figure 17: The double logarithmic diagrams of fan-out of DEP_S invocation for classes at all layers of a package.

4 4 5
4
3 3
GEN_all(i)

GEN_all(i)
GEN_all(i)

3
2 2
2
1 1
1
0 0 0

0 100 200 300 400 500 –100 0 100 200 300 400 500 600 700 0 200 400 600 800 1000
i i i
(a) (b) (c)

Figure 18: The double logarithmic diagrams of fan-out of GEN invocation for classes at all layers of a package.

Table 6: Values of fan-out for different classes of GEN.

Value of fan-out 0 1 2 3 4 5 7
Illusion 244 199 21 5 1 0 0
GEN_layer JabRef 479 192 16 6 1 0 0
GanttProject 630 264 45 6 0 1 0
Illusion 173 246 29 19 3 0 0
GEN_all JabRef 309 276 79 27 3 0 0
GanttProject 485 349 79 21 10 1 1

Table 7: Values of R, SD, and B parameters.

Software system R SD B
Illusion 0.9772 0.08465 2.51614
ASS_layer JabRef 0.91651 0.17962 7.1483
GanttProject 0.94567 0.11961 7.22368
Illusion 0.97249 0.31761 6.07342
DEP_D_layer JabRef 0.97537 0.12677 20.06268
GanttProject 0.95689 0.14931 25.12239
Illusion 0.95388 0.14533 5.52632
DEP_S_layer JabRef 0.94658 0.19504 14.97997
GanttProject 0.94086 0.1309 17.20075
Illusion 0.97414 0.10346 1.84955
ASS_all JabRef 0.88148 0.2079 1.65033
GanttProject 0.94905 0.13485 3.64204
Illusion 0.97028 0.14556 3.39056
DEP_D_all JabRef 0.97314 0.13056 8.47787
GanttProject 0.96602 0.13634 9.56684
Illusion 0.97265 0.1208 3.5066
DEP_S_all JabRef 0.96831 0.14207 4.47638
GanttProject 0.95419 0.12876 5.99337
Mathematical Problems in Engineering 15

between classes in software systems is significant. Based on 0, 1, and 2 for class inheritance in generalization, interface
data analysis, it can be seen that scale-free” property of implementation, and implementation of abstract classes, which
complex networks motivated software developers to pay more were approximately linearly distributed. Therefore, the power-
attention to the distribution range of the coupling in large- law value of GEN was approximated to 1.
scale software systems, which could provide a reliable ref-
erence for developing more reasonable software systems.
3.3.1. Calculation of Coupling Measurement for Art of
Illusion. According to the CSBG method for coupling
3.3. Coupling Measurement for the Three Open-Source Soft-
measurement, coupling of the software system for Art of
ware Systems. According to the results of the above-
Illusion was calculated as follows:
mentioned analysis, out-degrees of classes were often equal to

4 T
T bi 4 bi 4 bi 4 bi ⎤⎦
􏼂x1 , x2 , x3 , x4 􏼃 � ⎡⎣􏽘 ,􏽘 ,􏽘 ,􏽘
b b b b
i�1 1 i�1 2 i�1 3 i�1 4

4 4 4 4 T
bi bi bi b
� ⎡⎣􏽘 ,􏽘 ,􏽘 , 􏽘 i ⎤⎦
i�1
2.51614 i�1
6.07342 i�1
5.52632 i�1
1

� [6.007567, 2.48858, 2.735252, 15.11588]T ,

8 T
T bi 8 bi 8 bi 8 bi ⎤⎦
􏼂x5 , x6 , x7 , x8 􏼃 � ⎡⎣􏽘 ,􏽘 ,􏽘 ,􏽘
b b b b
i�5 5 i�5 6 i�5 7 i�5 8

8 8 8 8 T
bi bi bi b
� ⎡⎣􏽘 ,􏽘 ,􏽘 , 􏽘 i ⎤⎦
i�5
1.84955 i�5
3.39056 i�5
3.5066 i�5
1

� [5.269774, 2.874661, 2.779533, 9.74671]T ,

􏽐8i�4 bi
α� � 0.392023,
􏽐8i�1 bi

β � 1 − α � 0.607977,
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 (11)
􏼌􏼌 􏼌 􏼌 􏼌 􏼌 􏼌􏼌 􏼌􏼌 􏼌
􏼌􏼌􏽘 S layer(i)􏼌􏼌􏼌 � ⎡⎣􏼌􏼌􏼌􏽘 ASS layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP D layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP S layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 GEN layer(i)􏼌􏼌􏼌⎤⎦ × 􏼂x , x , x , x 􏼃T
􏼌􏼌 􏼌 􏼌􏼌
􏼌 􏼌 􏼌􏼌
􏼌 􏼌 􏼌􏼌
􏼌 􏼌 􏼌􏼌
􏼌 􏼌􏼌􏼌 1 2 3 4
􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1

� [2479, 5848, 2005, 260] ×[6.007567, 2.48858, 2.735252, 15.11588]T

� 38861.91,
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
􏼌􏼌 􏼌 􏼌 􏼌 􏼌 􏼌􏼌 􏼌 􏼌 􏼌
􏼌􏼌􏽘 S all(i)􏼌􏼌􏼌 � ⎡⎣􏼌􏼌􏼌􏽘 ASS all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP D all(i)(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP S all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 GEN all(i)􏼌􏼌􏼌⎤⎦ × 􏼂x , x , x , x 􏼃T
􏼌􏼌 􏼌 􏼌􏼌
􏼌 􏼌 􏼌􏼌
􏼌 􏼌 􏼌􏼌
􏼌 􏼌 􏼌􏼌
􏼌 􏼌􏼌􏼌 5 6 7 8
􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1

� [6705, 13883, 6255, 373] ×[5.269774, 2.874661, 2.779533, 9.74671]T

� 96264.25,
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌
􏼌 􏼌 􏼌
S � 􏼌􏼌􏼌􏽘 S(i)􏼌􏼌􏼌 � ⎡⎣􏼌􏼌􏼌􏽘 S layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 S all(i)􏼌􏼌􏼌􏼌⎤⎦ ×[α, β]T
􏼌 􏼌
􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌

� [38861.91, 96264.25] ×[0.392023, 0.607977]T

� 73761.21,

S 73761.21
S� � � 156.9387.
n 470
16 Mathematical Problems in Engineering

3.3.2. Calculation of Coupling Measurement for JabRef.

According to CSBG for coupling measurement, coupling of
the software system for JabRef was calculated as follows:

4 4 4 4 T
T ⎣􏽘 bi bi , 􏽘 bi , 􏽘 bi , 􏽘 bi ⎤⎦
􏼂 x1 , x 2 , x 3 , x 4 􏼃 � ⎡
b b b b
i�1 1 1 i�1 2 i�1 3 i�1 4
b

4 4 4 4 T
⎣􏽘 bi bi bi bi b
�⎡ ,􏽘 ,􏽘 , 􏽘 i ⎤⎦
i�1
7.1483 7.1483 i�1 20.06268 i�1 14.97997 i�1 1

� [6.042129, 2.152801, 2.883247, 43.19095]T ,

8 8 8 8 T
T ⎣􏽘 bi , 􏽘 bi , 􏽘 bi , 􏽘 bi ⎤⎦
􏼂 x5 , x 6 , x 7 , x 8 􏼃 � ⎡
b b b b
i�5 5 i�5 6 i�5 7 i�5 8

8 8 8 8 T
bi bi bi b
�⎣
⎡􏽘 ,􏽘 ,􏽘 , 􏽘 i ⎦⎤
i�5
1.65033 i�5 8.47787 i�5 4.47638 i�5 1

� [9.45543, 1.840625, 3.485982, 15.60458]T ,

􏽐8i�4 bi
α� � 0.26540419,
􏽐8i�1 bi

β � 1 − α � 0.734596,
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 (12)
􏼌􏼌 􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌
􏼌􏼌􏽘 S layer(i)􏼌􏼌􏼌 � ⎡ 􏼌 􏼌 􏼌 􏼌 􏼌 􏼌 􏼌 􏼌 T
􏼌􏼌 􏼌 ⎣􏼌􏼌􏼌􏽘 ASS layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP D layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP S all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 GEN layer(i)􏼌􏼌􏼌⎤⎦ × 􏼂x1 , x2 , x3 , x4 􏼃
􏼌
􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌

� [1326, 2032, 2396, 246] ×[6.042129, 2.152801, 2.883247, 43.19095]T

� 29375.8,
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
􏼌􏼌 􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌
􏼌􏼌􏽘 S all(i)􏼌􏼌􏼌 � ⎡
⎣ 􏼌􏼌􏽘 ASS all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP D all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP S all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 GEN all(i)􏼌􏼌􏼌⎤⎦ × 􏼂x , x , x , x 􏼃T
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 5 6 7 8
􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌

� [3430, 4403, 6733, 527] ×[9.45543, 1.840625, 3.485982, 15.60458]T

� 71947.46,
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌
􏼌 􏼌 ⎣􏼌􏼌􏽘 S layer(i)􏼌􏼌, 􏼌􏼌􏽘 S all(i)􏼌􏼌􏼌􏼌⎤⎦ ×[α, β]T
S � 􏼌􏼌􏼌􏽘 S(i)􏼌􏼌􏼌 � ⎡ 􏼌 􏼌 􏼌
􏼌 i�1 􏼌 􏼌􏼌 i�1 􏼌􏼌 􏼌􏼌
i�1
􏼌􏼌

� [29375.8, 71947.46] ×[0.26540419, 0.734596]T

� 60648.76,

S 60648.76
S� � � 87.39015.
n 694
Mathematical Problems in Engineering 17

3.3.3. Calculation of Coupling Measurement for GanttProject.

According to CSBG for coupling measurement, coupling of
the software system for GanttProject was calculated as
follows:

4 4 4 4 T
T ⎣􏽘 bi , 􏽘 bi , 􏽘 bi , 􏽘 bi ⎤⎦
􏼂 x1 , x2 , x3 , x4 􏼃 � ⎡
b b b b
i�1 1 i�1 2 i�1 3 i�1 4

4 4 4 4 T
⎣􏽘 bi bi bi b
�⎡ ,􏽘 ,􏽘 , 􏽘 i ⎤⎦
i�1
7.22368 i�1
25.12239 i�1
17.20075 i�1
1

� [6.997378, 2.012023, 2.93864, 50.54682]T ,

8 T
T ⎣ bi 8 bi 8 bi 8 bi ⎤⎦
􏼂 x5 , x 6 , x 7 , x 8 􏼃 � ⎡􏽘 ,􏽘 ,􏽘 ,􏽘
b b b b
i�5 5 i�5 6 i�5 7 i�5 8

8 8 8 8 T
bi bi bi b
�⎣
⎡􏽘 ,􏽘 ,􏽘 , 􏽘 i ⎦⎤
i�5
3.64204 i�5 9.56684 i�5 5.99337 i�5 1

� [5.5496, 2.111695, 3.370766, 20.20225]T ,

􏽐8i�4 bi
α� � 0.285548,
􏽐8i�1 bi

β � 1 − α � 0.714452,
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 (13)
􏼌􏼌 􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌 􏼌 􏼌
􏼌􏼌􏽘 S layer(i)􏼌􏼌􏼌 � ⎡
⎣􏼌􏼌􏽘 ASS layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP D layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP S layer(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 GEN layer(i)􏼌􏼌􏼌⎤⎦ × 􏼂x , x , x , x 􏼃T
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 1 2 3 4
􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌

� [1102, 1331, 565, 379] ×[6.997378, 2.012023, 2.93864, 50.54682]T

� 31206.69,
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
􏼌􏼌 􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌
􏼌􏼌􏽘 S all(i)􏼌􏼌􏼌 � ⎡
⎣􏼌􏼌􏽘 ASS all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP D all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 DEP S all(i)􏼌􏼌􏼌, 􏼌􏼌􏼌􏽘 GEN all(i)􏼌􏼌􏼌⎤⎦ × 􏼂x , x , x , x 􏼃T
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 5 6 7 8
􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌

� [3201, 3833, 2291, 622] ×[5.5491, 2.111695, 3.370766, 20.20225]T

� 46138.17,
􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌 􏼌􏼌 n 􏼌􏼌
􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌 􏼌􏼌
S � 􏼌􏼌􏽘 S(i)􏼌􏼌 � 􏼌􏼌􏽘 S layer(i)􏼌􏼌, 􏼌􏼌􏽘 S all(i)􏼌􏼌􏼌􏼌⎤⎦ ×[α, β]T
􏼌
􏼌 􏼌
􏼌 ⎣
⎡􏼌
􏼌 􏼌
􏼌 􏼌
􏼌
􏼌 i�1 􏼌 􏼌 i�1 􏼌 􏼌 i�1 􏼌

� [31206.69, 46138.17] ×[0.285548, 0.714452]T

� 41874.52,

S 41874.52
S� � � 44.26482.
n 946
18 Mathematical Problems in Engineering

The three aforementioned open-source software systems software systems using statistics of complex networks. In
were analyzed from the points of view of package level, class [54], it was mentioned that cohesion distribution of the
level, and method level using CSBG for coupling mea- majority classes of a software system contained a certain
surement. A program was also developed to calculate various regularity. In other words, it was not the case that neither
metrics; thus, the coupling of the three open-source software cohesion of all classes was very large nor very small. In the
systems in descending order was the Art of Illusion, JabRef, empirical analysis of coupling, the values of coupling metrics
and GanttProject, suggesting that it was feasible to use CSBG had a regularity similar to class cohesion. Although coupling
for coupling measurement of software systems that con- represented the degree of interdependence between classes,
tained a certain practical value. the greater the coupling, the more complex the software
from an intuitive aspect. However, excessive pursuit of “high
4. Conclusion cohesion and low coupling” of software systems increases
the workload of software developers and the complexity of
Based on bipartite graphs for complex networks, by com- software systems as well. Therefore, the empirical evidence
prehensive consideration of the weighted fan-out between showed that within a certain range, reducing the coupling
classes from points of view of package level, class level, and was helpful to attenuate the complexity of the software, while
method level, this study expressed that the interaction of excessively blindly pursuit of low coupling increases the
classes is a special bipartite graph, while a software system is complexity of software systems.
a set of these special bipartite graphs. For this purpose, first,
this study analyzed the four relationships for a software Data Availability
system, including ASS, DEP_D, DEP_S, and GEN, and
coupling relationship for a class with other classes in the The data used to support the findings of this study are
same layer of package was considered as well. Moreover, available from the corresponding author upon request.
coupling relationship for classes in a package with other
classes in different layers of the package was taken into Conflicts of Interest
account. Therefore, the CSBG method for coupling mea-
surement of software systems was proposed, which was The authors declare that there are no conflicts of interest
completely in compliance with the mathematical charac- regarding the publication of this paper.
teristics of the widely accepted metrics. Second, for a soft-
ware system, other typical methods and CSBG method were Acknowledgments
compared for the purpose of coupling measurement, and the This work was supported by the National Natural Science
results revealed that the measured value was either large or Foundation of China (no. 61602400). This work was also
small due to the defects of other measurement methods that supported in part by the Key Research and Development
were analyzed from an overall and global perspective. Program of Hangzhou under Grant 20182011A46.
Moreover, the corresponding values were mostly equal to 0.
Therefore, there were some defects in other measurement
References
methods, while CSBG had its rationality. Eventually, a
program was developed based on the CSBG method to apply [1] W. P. Stevens, G. J. Myers, and L. L. Constantine, “Structured
the three open-source software systems (Art of Illusion, design,” IBM Systems Journal, vol. 13, no. 2, pp. 115–139, 1974.
JabRef, and GanttProject). The results demonstrated that [2] W. P. Stevens, G. J. Myers, and L. L. Constantine, “Structured
coupling of the three open-source software systems in the design,” IBM Systems Journal, vol. 38, no. 2, pp. 231–256, 1999.
descending order was the Art of Illusion, JabRef, and [3] E. Yourdon and L. L. Constantine, Structured Design: Fun-
GanttProject. Although inheritance between classes in- damentals of a Discipline of Computer Program and Systems
Design, Yourdon Press, Englewood Cliffs, NJ, USA, 1979.
creases coupling of the system, this is also followed by
[4] L. C. Briand, J. W. Daly, and J. K. Wust, “A unified framework
software engineering, which is conducive to reduce function for coupling measurement in object-oriented systems,” IEEE
definition and attribute definition in order to create a new Transactions on Software Engineering, vol. 25, no. 1, pp. 91–
class, and thus, this is weak coupling. It can be concluded 121, 1999.
from the linear distribution of GEN that all classes either had [5] M. D’Ambros, M. Lanza, and R. Robbes, “On the relationship
an inheritance relationship, or that the number of GEN fan- between change coupling and software defects,” in Proceed-
out of all classes was very large or very small. However, ings of the 2009 16th Working Conference on Reverse
classes with values equal to 0 or 1 were accounted. Fur- Engineering, Lille, France, October 2009.
thermore, it was revealed that in the same layer and total [6] T. Gyimothy, R. Ferenc, and I. Siket, “Empirical validation of
layers of the package, fan-out values of ASS, DEP_D, and object-oriented metrics on open source software for fault
DEP_S obeyed the scale-free property of complex networks. prediction,” IEEE Transactions on Software Engineering,
vol. 31, no. 10, pp. 897–910, 2005.
These findings provided empirical support for the CSBG
[7] P. Yu, T. Systa, and H. Muller, “Predicting fault-proneness
method. The statistical power-law metrics were applied to using OO metrics. An industrial case study,” in Proceedings of
the method for coupling measurement proposed in this the Sixth European Conference on Software Maintenance and
study in order to calculate the coupling of the three open- Reengineering, Budapest, Hungary, March 2002.
source software systems, which provided a reliable reference [8] L. C. Briand, J. Wuest, and H. Lounis, “Using coupling
for further investigation of coupling between classes in measurement for impact analysis in object-oriented systems,”
Mathematical Problems in Engineering 19

in Proceedings of the IEEE International Conference on Soft- [26] L. C. Briand, S. Morasca, and V. R. Basili, “Property-based
ware Maintenance, Oxford, UK, September 1999. software engineering measurement,” IEEE Transactions on
[9] F. G. Wilkie and B. A. Kitchenham, “Coupling measures and Software Engineering, vol. 22, no. 1, pp. 68–86, 1996.
change ripples in C++ application software,” Journal of [27] L. C. Briand, J. W. Daly, and J. Wüst, “A unified framework
Systems & Software, vol. 52, no. 2-3, pp. 157–164, 2000. for cohesion measurement in object-oriented systems,” Em-
[10] G. Antoniol, R. Fiutem, and L. Cristoforetti, “Using metrics to pirical Software Engineering, vol. 3, no. 1, pp. 65–117, 1998.
identify design patterns in object-oriented software,” in [28] D. J. Watts and S. H. Strogatz, “Collective dynamics of “small-
Proceedings of the Fifth International Software Metrics world” networks,” Nature, vol. 393, 1998.
Symposium, Bethesda, MD, USA, November 1998. [29] A.-L. Barabási and R. Albert, “Emergence of scaling in ran-
[11] S. R. Chidamber and C. F. Kemerer, “A metrics suite for object dom networks,” Science, vol. 286, no. 5439, pp. 509–512, 1999.
oriented design,” IEEE Transactions on Software Engineering, [30] W. Pan, H. Ming, C. K. Chang, Z. Yang, and D.-K. Kim,
vol. 20, no. 6, pp. 476–493, 1994. “ElementRank: ranking java software classes and packages
[12] S. R. Chidamber and C. F. Kemerer, “Towards a metrics suite using multilayer complex network-based approach,” IEEE
for object oriented design,” in Proceedings of the ACM Transactions on Software Engineering, 2019.
Conference on Object Oriented Programming, Systems, Lan- [31] W. Pan, B. Li, J. Liu, Y. Ma, and B. Hu, “Analyzing the
guages and Applications, Orlando, FL, USA, 1991. structure of java software systems by weighted K-core de-
[13] W. Li and S. Henry, “Object-oriented metrics that predict composition,” Future Generation Computer Systems, vol. 83,
maintainability,” Journal of Systems and Software, vol. 23, pp. 431–444, 2018.
no. 2, pp. 111–122, 1993. [32] C. R. Myers, “Software systems as complex networks:
[14] Y. Lee, “Measuring the coupling and cohesion of an object- structure, function, and evolvability of software collaboration
oriented program based on information flow,” in Proceedings graphs,” Physical Review E, vol. 68, no. 4, 15 pages, 2003.
of the International Conference on Software Qualiy, Maribor, [33] N. LaBelle and E. Wallingford, “Inter package dependency
Slovenia, 1995. networks in open source software,” 2004, http://arxiv.org/abs/
[15] L. Briand, P. Devanbu, and W. Melo, “An investigation into 0411096.
coupling measures for C++ software engineering,” in Pro- [34] D. Hyland-Wood, D. Carrington, and S. Kaplan, “Scale-free
ceedings of the 19th International Conference on Software nature of java software package,” class and method collabo-
ration graphs,” Techical report no. TR-MS1286, University of
engineering, Boston, MA, USA, May 1997.
[16] H. Li and B. Li, “A pair of coupling metrics for software Maryland College, College Park, MD, USA, 2006.
[35] Y. Xiang, W. Pan, H. Jiang, Y. Zhu, and H. Li, “Measuring
networks,” Journal of Systems Science and Complexity, vol. 24,
software modularity based on software networks,” Entropy,
no. 1, pp. 51–60, 2011.
vol. 21, no. 4, p. 344, 2019.
[17] E. Arisholm, L. C. Briand, and A. Foyen, Dynamic Coupling
[36] W. Pan and C. Chai, “Measuring software stability based on
Measurement for Object-Oriented Software, IEEE Press, Pis-
complex networks in software,” Cluster Computing, vol. 22,
cataway, NJ, USA, 2004.
no. S2, pp. 2589–2598, 2019.
[18] J. K. Chhabra and V. Gupta, “A survey of dynamic software
[37] W. Pan, B. Song, K. Li, and K. Zhang, “Identifying key classes
metrics,” Journal of Computer Science & Technology, vol. 25,
in object-oriented software using generalizedk-core decom-
no. 5, pp. 1016–1029, 2010.
position,” Future Generation Computer Systems, vol. 81,
[19] D. Poshyvanyk and A. Marcus, “The conceptual coupling
pp. 188–202, 2018.
metrics for object-oriented systems,” in Proceedings of the [38] W. Pan and C. Chai, “Structure-aware mashup service
2006 22nd IEEE International Conference on Software clustering for cloud-based internet of things using genetic
Maintenance, Philadelphia, PA; USA, September 2006. algorithm based clustering algorithm,” Future Generation
[20] D. Poshyvanyk, A. Marcus, R. Ferenc, and T. Gyimóthy, Computer Systems, vol. 87, pp. 267–277, 2018.
“Using information retrieval based coupling measures for [39] W. Pan, J. Dong, K. Liu, and J. Wang, “Topology and topic-
impact analysis,” Empirical Software Engineering, vol. 14, aware service clustering,” International Journal of Web Ser-
no. 1, pp. 5–32, 2009. vices Research, vol. 15, no. 3, pp. 18–37, 2018.
[21] M. Gethers and D. Poshyvanyk, “Using relational topic [40] F. Liljeros, C. R. Edling, L. A. N. Amaral, H. E. Stanley, and
models to capture coupling among classes in object-oriented Y. Åberg, “The web of human sexual contacts,” Nature,
software systems,” in Proceedings of the 2010 IEEE Interna- vol. 411, no. 6840, pp. 907-908, 2001.
tional Conference on Software Maintenance, Timisoara, [41] H. Jeong, B. Tombor, R. Albert, Z. N. Oltvai, and
Romania, September 2010. A.-L. Barabási, “The large-scale organization of metabolic
[22] H. Gall, K. Hajek, and M. Jazayeri, “Detection of logical networks,” Nature, vol. 407, no. 6804, pp. 651–654, 2002.
coupling based on product release history,” in Proceedings of [42] P.-P. Zhang, K. Kan Chen, Y. He et al., “Model and empirical
the International Conference on Software Maintenance, study on some collaboration networks,” Physica A: Statistical
Washington, DC, USA, April 1998. Mechanics and Its Applications, vol. 360, no. 2, pp. 599–616,
[23] T. Zimmermann, A. Zeller, P. Weissgerber, and S. Diehl, 2006.
“Mining version histories to guide software changes,” IEEE [43] Q. Xuan, F. Du, and T. J. Wu, “Empirical analysis of internet
Transactions on Software Engineering, vol. 31, no. 6, telephone network : from user ID to phone,” Chaos: An In-
pp. 429–445, 2005. terdisciplinary Journal of Nonlinear Science, vol. 19, no. 2,
[24] E. J. Weyuker, “Evaluating software complexity measures,” Article ID 023101, 2009.
IEEE Transactions on Software Engineering, vol. 14, no. 9, [44] M.-S. Shang, L. Lü, Y.-C. Zhang, and T. Zhou, “Empirical
pp. 1357–1365, 1988. analysis of web-based user-object bipartite networks,” EPL
[25] I. Vessey and R. Weber, “Research on structured program- (Europhysics Letters), vol. 90, no. 4, p. 48006, 2010.
ming: an empiricist’s evaluation,” IEEE Transactions on [45] K.-I. Goh, M. E. Cusick, D. Valle, B. Childs, M. Vidal, and
Software Engineering, vol. SE-10, no. 4, pp. 397–407, 2009. A.-L. Barabasi, “The human disease network,” Proceedings of
20 Mathematical Problems in Engineering

the National Academy of Sciences, vol. 104, no. 21,

pp. 8685–8690, 2007.
[46] B. Kitchenham, “What’s up with software metrics?-a pre-
liminary mapping study,” Journal of Systems and Software,
vol. 83, no. 1, pp. 37–51, 2010.
[47] J. Eder and M. Schrefl, “Coupling and cohesion in object-
oriented systems,” in Proceedings of the International
Workshop on Object Orientation in Operating Systems,
pp. 264–272, Dordan, France, September 1992.
[48] Illusion, 2012, http://sourceforge.net/projects/aoi/.
[49] JabRef, 2012, http://sourceforge.net/projects/jabref/.
[50] GanttProject, 2012, http://sourceforge.net/projects/ganttproject/.
[51] J. Al Dallal and L. C. Briand, “An object-oriented high-level
design-based class cohesion metric,” Information and Soft-
ware Technology, vol. 52, no. 12, pp. 1346–1361, 2010.
[52] J. Al Dallal, “Measuring the discriminative power of object-
oriented class cohesion metrics,” IEEE Transactions on Soft-
ware Engineering, vol. 37, pp. 778–804, 2011.
[53] J. Al Dallal, “The impact of accounting for special methods in
the measurement of object-oriented class cohesion on
refactoring and fault prediction activities,” Journal of Systems
and Software, vol. 85, no. 5, pp. 1042–1057, 2012.
[54] A. Gu, X. Zhou, Z. Li, Q. Li, and L. Li, “Measuring object-
oriented class cohesion based on complex networks,” Arabian
Journal for Science and Engineering, vol. 42, no. 8, pp. 3551–
3561, 2017.