Hindawi Publishing Corporation

Discrete Dynamics in Nature and Society

Volume 2016, Article ID 4643630, 15 pages
http://dx.doi.org/10.1155/2016/4643630

Research Article
Bidding Strategy to Support Decision-Making Based on
Comprehensive Information in Construction Projects

Ru Liang,1 Zhaohan Sheng,1 Feng Xu,1 and Changzhi Wu2

1
School of Management and Engineering, Nanjing University, Nanjing, Jiangsu 210093, China
2
Australasian Joint Research Centre for Building Information Modelling, School of Built Environment,
Curtin University, Bentley, WA 6102, Australia

Correspondence should be addressed to Ru Liang; rliang124@163.com

Received 27 May 2016; Accepted 29 June 2016

Academic Editor: Yong Zhou

Copyright © 2016 Ru Liang 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.

This paper develops a unified method to support contractor for bidding selection in construction projects. A cross-functional
contractor with 28 candidate units distributed in the three departments (construction units, design units, and suppliers) is used as
an example. This problem is first formulated as a 0-1 quadratic programming problem through optimizing individual performance
and collaborative performance of the candidate units based on individual information and collaborative information. Then, a
multiobjective evolutionary algorithm is designed to solve this problem and a bidding selection problem for a major bridge project
is used to demonstrate our proposed method. The results show that the decision-maker (DM) obtains a better contractor if he pays
more attention to collaborative performance.

1. Introduction frequently stated that “the resulting fierce competition for jobs
forces construction companies to look for more sophisticated
In bidding process of construction projects, contractors have analytical tools to analyze and improve their bidding strategies”
to make several simultaneous and interdependent decisions [9]. This leads to the conclusion that “[construction] managers
despite the limited information available [1]. Numerous need statistical estimation techniques for effectively mining data
studies are concerned with bidding strategies in construction generated by auctions to predict future behavior and to dynam-
projects [2]. The drawback of these kinds of strategies is ically improve operational decisions” [10]. One approach of
that it requires specific information about their competitors. acquiring competitiveness information is to monitor past
Furthermore, many of them are developed for estimating bidding behavior, but this is subjectively based in the con-
construction costs rather than estimating bid amounts [3, struction setting [6]. Nowadays, innovative approaches to
4]. In practice, bidding decision of construction projects is procurement include sequential markets, dynamic bidding
a complex decision-making process affected by numerous models, combinatorial auctions, online auctions, and mar-
factors, such as site terrain, environmental conditions, con- ket places [6]. Existing conceptual frameworks on bidding
struction method, climate conditions, and funding ability [5]. selection are developed through scattered efforts in more
Bidding performance concerns the relationship among general construction contexts for assessing and comparing
different bidders whose bids are submitted in a competition performance of construction companies [11]. Frameworks to
[6]. Currently, internationalization of construction compa- address bidding performance are rather sparse since most
nies has attracted broad attention and significant research of them are focusing on bidding accuracy, namely, cost
interest because of global economic slowdown and construc- estimating accuracy [12]. Moreover, it is paradoxical since lots
tion demand shrinkage [7]. Similarly, predictive information of research concerning competitiveness in bidding has been
concerns the competitiveness of contractors which is a published linking the size of the bidder and the size of the
potentially valuable asset for multiple DMs involved in the contracts, that is, proving that there are usually some affinities
construction procurement process [6, 8]. For instance, it is between them [13].
2 Discrete Dynamics in Nature and Society

A bidding process usually requires adoption of multi- In this paper, we will study contractor selection of bidding
phase strategies and criteria [14, 15]. For instance, Bagies and in construction projects based on individual information
Fortune [15] divide a bidding process into four main cate- and collaborative information. A 0-1 quadratic programming
gories: bidding strategies, bidding markup determinations, model of the cross-functional contractors on bidding in
factors affecting bidding decisions, and bidding decision- construction projects based on individual information and
making process. Meanwhile, identifying the main factors has collaborative information is established. The results show that
been studied on project factors [5, 16]. In addition, several DMs obtain the optimal contractor if they pay more attention
multicriteria bidding models [17] focus on considering vari- to collaborative performance. The major contribution of our
ous factors to achieve optimal markup decision for construc- method is twofold: (1) The model takes into account different
tion projects [18, 19]. However, they usually use an additive numbers of design units, construction units, and suppliers.
utility approach through ignoring possible interdependence We not only optimize individual performance of design
between different bidding criteria [1]. Moreover, many studies units, construction units, and suppliers, but also optimize
focus on estimating a successful bidding by effective methods their collaborative performance. Moreover, our model can
based on project factors [14, 20, 21]. For instance, Wanous be embedded into decision support system of construction
et al. [14] identify 35 factors affecting bidding decision, projects to deal with large scale and complex contractor
formulate a linear equation based on a staff survey, and selection of bidding based on individual information and
propose an assessment model and neural score to determine collaborative information. (2) A novel multiobjective evolu-
whether a bidder is advantageous or not. Most of the existing tionary algorithm for solving the 0-1 quadratic programming
studies focus on identifying relevant factors and dealing with model of the cross-functional contractor is presented. In this
the correlation among factors and bidding results without real method, a new strategy is presented to handle congestion
case studies. among the individuals with simple computation. Based on
Inspired by the abovementioned studies which high- this algorithm, we can obtain a set of Pareto optimal solutions.
lighted key attributes that form the basis of the bidding Then, DMs can select design units, construction units, and
decision-making, various approaches are developed for the the suppliers according to individual performance and col-
contractor to select bidding in construction projects, such as laborative performance to form a cross-functional contractor.
the integrated multiobjective decision-making process [22], The rest of this paper is organized as follows. Section 2 states
the neural networks [23], AHP [24, 25], data envelopment contractor selection problem of bidding in construction
analysis [26], the multiattribute analysis model [27], analytic projects based on individual information and collaborative
network process [28], integer programming [29], multiat- information. Then, a general description framework of the
tribute utility theory [30], and the binary goal programming problem is given. Section 3 formulates this problem as a
model [31]. These models have insightful implications for 0-1 multiobjective optimization problem, and an algorithm
the management side of a construction firm and reveal is presented to solve the problem in Section 4. Section 5
important information about bidding practices and patterns illustrates the proposed method through a detailed example
in different construction markets [32]. These approaches have and Section 6 concludes the paper.
drawn much attention to the area of contractor selection
in construction projects and provided basis and new ideas 2. Problem Description
to group decision-making [32]. In addition, group decision-
making models are applied in outsourcing contracts selection In this section, we will describe how a contractor selects
[33], selecting project team members and suppliers [34], and bidding in construction projects based on individual infor-
selecting repair contracts [35] in the area of construction mation and collaborative information. Construction projects
projects. Then, some scholars [36, 37] address the contribu- are typically user-driven, which allocate tasks step by step
tions of experts from different professions to group decision- through contractors and absorb units from different depart-
making for selecting contractors of bidding in construction ments (organizations) in the network structures. A contrac-
projects. For instance, Liu et al. [32] establish a two-stage tor management system of a construction project can be
partial least square path modeling aggregation approach described as multiagents, including owners, design contrac-
to expand the research area of group decision-making in tors, construction contractors, contractors, and equipment
large scale construction projects, with different direction suppliers. Meanwhile, they are also supply-demand systems
correlations taken into consideration among the evaluation of construction projects including contracts, suppliers, infor-
indicators. mation, and responsibility. In the following, construction
Existing research shows that the probabilistic risk analysis units, design units, and suppliers are taken as an example to
is superior to most risk analysis methods. However, this illustrate how to form a cross-functional contractor, namely,
may not be true since probabilistic risk analysis requires contractor selection of bidding in construction projects.
adequate and precise data from similar projects implemented As mentioned above, this paper solves contractor
in quite similar environments and working conditions to selection of bidding in construction projects to form a
get probability density functions. In practice, collecting such cross-functional contractor from different departments. For
information seems almost impossible because of the unique- instance, construction units, design units, and suppliers
ness of each construction project and unique features of every come from department 1, department 2, and department
certain contract. Thus, estimations of these experts may be 3, respectively. We consider both individual information
the most useful and dependable information in the cases. and collaborative information of the candidate units.
Discrete Dynamics in Nature and Society 3

Construction unit 10
Construction unit 12

Supplier 4 Supplier 6
Supplier 3 Design unit 2
Construction unit 11
Supplier 5
Construction unit 7 Construction unit 8
Construction unit 9
Construction unit 4

Supplier 9 Design unit 5 Construction unit 6

Design unit 1 Design unit 3 Supplier 7
Supplier 10
Supplier 8

Construction unit 2 Construction unit 5

Design unit 4 Design unit 6

Construction unit 1

Supplier 2 Construction unit 3 Supplier 1

Figure 1: Collaborative network constitutes 28 candidate units distributed in 3 departments.

Individual information refers to special characteristics of estimated indexes on individual performance. In order to
individual candidate units, such as working experience, calculate expediently, this paper adopts common estimated
qualification rate of completion inspection and acceptance, indexes of the three departments. On the other hand,
and professional knowledge. Collaborative information horizontal lines indicate the interaction among different
refers to historical relations of cooperation to complete tasks units. According to different departments, there are two kinds
in the past between two candidate units distributed in two of the interaction. If two units come from one department,
different departments. We select a contractor distributed in then the interaction between them refers to synergistic effect
three departments with 28 candidate units (design units, under estimated indexes of the individual performance.
construction units, and suppliers) shown in Figure 1 as an On the contrary, if two units come from two different
example, where pink circles denote design units; red circles departments, then the interaction between them refers to
indicate construction units; blue circles denote suppliers; and synergistic effect under estimated indexes of the collaborative
solid lines indicate interaction between two different units. performance. In addition, thicker horizontal lines have larger
Our target is to select one design unit from department 1, one synergistic effects.
construction unit from department 2, and one supplier from According to the problem described above, contractor
department 3, respectively. The distribution of the candidate selection of bidding in construction projects can be described
units can be described through a collaborative network based as in Figure 2. The aforementioned 28 candidate units (12
on interactions among the individual candidate units. construction units, 6 design units, and 10 suppliers) dis-
In Figure 1, there are 6 design units, 10 suppliers, and 12 tributed in three departments are used to illustrate the prob-
construction units from the 28 candidate units. On the one lem. Now, we explain four layers of the problem framework
hand, if circle areas of units are greater, the units have greater as follows:
individual performance than the other units from the same (1) First Layer. Overall goal of the problem is to form a
departments. For instance, because the circle area of design cross-functional contractor with optimizing individ-
unit 1 from department 1 is obviously larger than the other ual performance (quality factors, build capacity, deliv-
design units from department 1, the individual performance ery cycle, price, etc.) and collaborative performance
of design unit 1 is better than the others in department 1. (cooperation, credibility, etc.).
Similarly, construction units of department 2 and suppliers
of department 3 have the same characteristic as department (2) Second Layer. Under the overall goal, there are
1. However, it is worth mentioning that this feature is effective two subgoals including optimizing individual perfor-
only when the two units come from the same department. For mance and collaborative performance.
instance, the individual performance of supplier 10 of depart- (3) Third Layer. Individual performance of the can-
ment 3 cannot compare with the individual performance didate units is measured by individual indicators
of construction unit 3 from department 2, even though the 𝐼1 , 𝐼2 , . . . , 𝐼𝑚 . Meanwhile, collaborative performance
circle area of construction unit 3 is larger than the circle area of the candidate units is measured by collaborative
of supplier 10. Different departments usually have different indicators 𝐶1 , 𝐶2 , . . . , 𝐶𝑛 .
4 Discrete Dynamics in Nature and Society

Overall target Elements in the matrix 𝐴 = [𝑎𝑖𝑙 ]𝑛×𝑙 can be standardized in

the following way:
𝑎𝑖𝑙 − 𝑎𝑙min
𝑎𝑖𝑙󸀠 =
Individual Collaborative
, 𝑖 = 1, . . . , 𝑛; 𝑙 = 1, . . . , 𝑔, (1)
performance performance
𝑎𝑙max − 𝑎𝑙min

𝑎𝑙max − 𝑎𝑖𝑙
I1 I2 I3 ··· Im C1 C2 C3 · · · Cn 𝑎𝑖𝑙󸀠 = , 𝑖 = 1, . . . , 𝑛; 𝑙 = 1, . . . , 𝑔, (2)
𝑎𝑙max − 𝑎𝑙min

𝑎𝑙max = max {𝑎𝑖𝑙 | 𝑖 = 1, . . . , 𝑛} , 𝑙 = 1, . . . , 𝑔, (3)

𝑎𝑙min = min {𝑎𝑖𝑙 | 𝑖 = 1, . . . , 𝑛} , 𝑙 = 1, . . . , 𝑔. (4)

Assume that weight vector of individual indicators is 𝑈 =

(𝑢1 , 𝑢2 , . . . , 𝑢𝑔 ), which is obtained by the direct distribution
method and the AHP method. Then, we can obtain compre-
hensive individual performance of design units, construction
units, and suppliers through simple linear weighted method
as follows:
𝑔
Figure 2: A cross-functional contractor selection problem frame- 𝛼𝑖 = ∑𝑢𝑙 𝑎𝑖𝑙󸀠 , 𝑖 = 1, . . . , 𝑛. (5)
work. 𝑙=1

In the case of only considering individual performance,

(4) Fourth Layer. The candidate units which interact the performance of candidate unit 𝑝𝑖 is better if 𝛼𝑖 is greater.
with each other are distributed in three different According to comprehensive values 𝛼1 , 𝛼2 , . . . , 𝛼𝑛 of individ-
departments in the collaborative network. ual performance, we can obtain the following optimization
model:
3. Proposed Model 𝑛
Max 𝑍1 = ∑𝛼𝑖 𝑥𝑖 ,
Based on the problem described in Section 2, this section 𝑖=1
aims to establish a mathematical model for contractor selec- (6)
tion of bidding in construction projects based on individ- s.t. ∑ 𝑥𝑖 = 𝑞𝑗 , 𝑗 = 1, . . . , ℎ,
𝑖∈𝑁𝑗
ual information and collaborative information. Firstly, we
establish a selection model for construction units, design 𝑥𝑖 ∈ {0, 1} , 𝑖 = 1, . . . , 𝑛,
units, and suppliers based on individual information, which is
referred to as model 1. Then, a selection model of construction
where ∑ℎ𝑗=1 𝑞𝑗 = 𝑞.
units, design units, and suppliers based on collaborative
information is established, which is referred to as model 2. Model (6) is a binary optimization problem. It is to select 𝑞
Secondly, we integrate the two models and obtain an inte- units from 𝑛 candidate units (design units, construction units,
grated selection model for cross-functional contractors based and suppliers) and to maximize comprehensive values of
on individual information and collaborative information. The individual performance. Through descending comprehensive
process for choosing units from the three departments is values 𝛼1 , 𝛼2 , . . . , 𝛼𝑛 of individual performance, we can select
described in Figure 3. headmost 𝑞 units and obtain the optimal solution directly.

3.2. Selection Model Based on Collaborative Information. Let

3.1. Selection Model Based on Individual Information. Let
𝐷𝑚 = [𝑑𝑖𝑗𝑚 ]𝑛×𝑛 indicate collaborative decision matrix to
𝐴 = [𝑎𝑖𝑙 ]𝑛×𝑙 denote individual decision matrix to which 𝑎𝑖𝑙 is
performance on alternative unit 𝑝𝑖 with regard to individual which 𝑑𝑖𝑗𝑚 (𝑖 ≠ 𝑗) is performance on alternative unit 𝑝𝑖
indicator 𝐼𝑙 . Aiming at contractor selection of bidding in cooperating with alternative unit 𝑝𝑗 with regard to col-
construction projects, individual indicators include work laborative indicator 𝐶𝑚 . Aiming at contractor selection of
experience, work ability, and professional knowledge. There bidding in construction projects, collaborative indicators
are two kinds of individual indicators considered in this include cooperative research communication and sharing
paper. If individual indicator 𝐼𝑙 is an objective indicator, complementary knowledge base and harmonious culture. In
corresponding indicator value of 𝐼𝑙 is statistical data or general, imbalanced reciprocities are allowed among collabo-
measured data. If individual indicator 𝐼𝑙 is a subjective rative indicators [38]. For example, 𝑝𝑖 can get benefit from 𝑝𝑗 ,
indicator, corresponding indicator value of 𝐼𝑙 scores from 1 but 𝑝𝑗 will not contribute to 𝑝𝑖 . Thus, for some collaborative
to 10 (1: very poor, 10: very good) by experts. indicators, the performance of candidate unit 𝑝𝑖 coordinated
A standardized method is effective in measuring indicator with 𝑝𝑗 is different from the performance of candidate unit
values of design units, construction units, and suppliers. 𝑝𝑗 coordinated with 𝑝𝑖 , that is, 𝑑𝑖𝑗𝑚 ≠ 𝑑𝑗𝑖
𝑚
.
Discrete Dynamics in Nature and Society 5

Start
Optimal selection of
candidate units

Obtain bidding
information by
enterprise
Achieve No Reduce the target
selection target or ease conditions

Determine set of
Yes
candidate units
Calculate comprehensive
attribute values of
candidate units
Calculate individual
attribute values of
candidate units
Build units set of the
contractor

Calculate collaborative
attribute values of
candidate units

End

Figure 3: Building process for choosing units of the contractor.

Let 𝑑𝑖𝑖𝑚 = ‘−’ denote that collaborative information of Assume that weight vector for collaborative indicators
candidate unit 𝑝𝑖 is not to be considered; namely, diagonal is 𝑉 = (V1 , V2 , . . . , V𝑙 ), which could be obtained by direct
elements of matrix 𝐷𝑚 = [𝑑𝑖𝑗𝑚 ]𝑛×𝑛 are 0. It should be distribution method and AHP method. Then, we can obtain
pointed out that there are three ways of dealing with new comprehensive value of collaborative performance between
candidate units without collaborative information to the candidate units 𝑝𝑖 and 𝑝𝑗 through simple linear weighted
existing candidate units of the candidate set. The first way is method as follows:
to consider that collaborative information of new candidate 𝑘
units coordinated with the others is 0. The second way is to 𝛽𝑖𝑗 = ∑ V𝑚 𝑑𝑖𝑗󸀠𝑚 , 𝑖, 𝑗 = 1, . . . , 𝑛; 𝑖 ≠ 𝑗. (11)
consider that DMs consult cooperation information records 𝑚=1
of new candidate units in the other departments. The third
way is to consider that DMs give their expectations of future According to comprehensive values 𝛼1 , 𝛼2 , . . . , 𝛼𝑛 of col-
cooperation between new candidate units with the others laborative performance, we can obtain the following opti-
referring to cooperation information records between them. mization model:
Similarly, elements in the matrix 𝐷𝑚 = [𝑑𝑖𝑗𝑚 ]𝑛×𝑛 can be 𝑛 𝑛
standardized, where the calculating formulas are as follows: Max 𝑍2 = ∑ ∑ 𝛽𝑖𝑗 𝑥𝑖 𝑥𝑗 ,
𝑖=1 𝑗=1,𝑗=𝑖̸

𝑑𝑖𝑗𝑚 − 𝑑𝑚min s.t. ∑ 𝑥𝑖 = 𝑞𝑗 , 𝑗 = 1, . . . , ℎ, (12)

𝑑𝑖𝑗󸀠𝑚 = ,
𝑑𝑚max − 𝑑𝑚min (7) 𝑖∈𝑁𝑗

𝑖, 𝑗 = 1, . . . , 𝑛; 𝑖 ≠ 𝑗; 𝑚 = 1, . . . , 𝑘 (efficiency indicators) , 𝑥𝑖 ∈ {0, 1} , 𝑖 = 1, . . . , 𝑛,

𝑑𝑚max − 𝑑𝑖𝑗𝑚
𝑑𝑖𝑗󸀠𝑚 = , where ∑ℎ𝑗=1 𝑞𝑗 = 𝑞.
𝑑𝑚max − 𝑑𝑚min (8) In model (12), our target is to select 𝑞 from 𝑛 candidate
𝑖, 𝑗 = 1, . . . , 𝑛; 𝑖 ≠ 𝑗; 𝑚 = 1, . . . , 𝑘 (cost indicators) , units (design units, construction units, and suppliers) and
to maximize comprehensive values of collaborative perfor-
𝑑𝑚max = max {𝑑𝑖𝑗𝑚 | 𝑖, 𝑗 = 1, . . . , 𝑛; 𝑖 ≠ 𝑗} , 𝑚 = 1, . . . , 𝑘, (9) mance. It is observed that model (12) is a 0-1 quadratic
programming problem. In addition, model (12) is similar
𝑑𝑚min = min {𝑑𝑖𝑗𝑚 | 𝑖, 𝑗 = 1, . . . , 𝑛; 𝑖 ≠ 𝑗} , 𝑚 = 1, . . . , 𝑘. (10) to the model established by Kuo et al. [39] for biggest
6 Discrete Dynamics in Nature and Society

differentiation problem, which has been proved as an NP- Generate the initial
Binary coding
Hard problem [39]. population

3.3. Selection Model of Cross-Functional Contractor. In order

to solve contractor selection of bidding in construction Design and implement
projects based on individual and collaborative information, Double node cross
of crossover operator
we integrate model (6) and model (12) to the following 0-1
quadratic programming model:

𝑍 = 𝜆 Max 𝑍1 + 𝜇 Max 𝑍2 Reverse mutational Design of mutational

operation operator
𝑛 𝑛 𝑛
= 𝜆∑𝛼𝑖 𝑥𝑖 + 𝜇∑ ∑ 𝛽𝑖𝑗 𝑥𝑖 𝑥𝑗 ,
𝑖=1 𝑖=1 𝑗=1,𝑗=𝑖̸
Quick sort of Calculate individual
inferior quality fitness value
s.t. ∑ 𝑥𝑖 = 𝑞𝑗 , 𝑗 = 1, . . . , ℎ,
𝑖∈𝑁𝑗
(13)

𝑥𝑖 ∈ {0, 1} , 𝑖 = 1, . . . , 𝑛, Meeting No
terminational conditions?
0 ≤ 𝜆, 𝜇 ≤ 1,
Yes
𝜆 + 𝜇 = 1,
End
where ∑ℎ𝑗=1 𝑞𝑗 = 𝑞.
For model (13), its solution space is a function of 𝑛 and 𝑞.
Figure 4: Algorithm process for units selection model of the
Let Φ be number of solutions in the solution space; then the contractor.
number of possible solutions is

𝑞/ℎ ℎ
Φ = 𝐶𝑛𝑞11 × 𝐶𝑛𝑞22 × ⋅ ⋅ ⋅ × 𝐶𝑛𝑞ℎℎ ≤ [𝐶𝑛/ℎ ] 4. Model Solving
ℎ (14) Evolutionary algorithm for solving multiobjective optimiza-
(𝑛/ℎ)!
={ } . tion problems not only can obtain a one-time multiobjective
[(𝑛/ℎ) − (𝑞/ℎ)]! × (𝑞/ℎ)! optimization problem of multiple Pareto optimal solutions,
but also has good consistency of optimization results. MOEA
According to properties of combinations, we have 𝐶𝑛𝑞 = is such a kind of optimization search method, which mainly
𝐶𝑛𝑛−𝑞 .
When 𝑞 ≤ 𝑛/2, 𝑛 − 𝑞 ≤ 𝑛/2; and when 𝑞/ℎ ≪ 𝑛/ℎ, Φ includes genetic algorithm, genetic programming, and evolu-
could be handled approximately as tionary programming [39, 41, 42]. In addition, it has robust-
ness to species of optimization problems. More importantly,
𝑞/ℎ ℎ
Φ = 𝐶𝑛𝑞11 × 𝐶𝑛𝑞22 × ⋅ ⋅ ⋅ × 𝐶𝑛𝑞ℎℎ ≤ [𝐶𝑛/ℎ ] its optimized result is also a set for evolutionary operation
on population made up of many individuals. Therefore, it is
ℎ particularly suitable for solving complex NP-Hard problem
(𝑛/ℎ)!
={ } of multiobjective decision-making (optimization). Therefore,
[(𝑛/ℎ) − (𝑞/ℎ)]! × (𝑞/ℎ)!
due to characteristics of model (13), we design MOEA of the
(15)
𝑛 𝑛 𝑛 𝑞 ℎ model. The details of algorithm design process are shown in
≤ {( ) × [( ) − 1] × ⋅ ⋅ ⋅ × − [( ) − ( ) + 1]} Figure 4.
ℎ ℎ ℎ ℎ
ℎ
𝑛 𝑞/ℎ 𝑛 𝑞 4.1. Generate the Initial Population. Using binary coding,
≤ [( ) ] = ( ) .
ℎ ℎ namely, 0-1 code, we encode [0, 1, 0, . . . , 0, 1] for all con-
struction units, design units, and suppliers of bidding. The
Through the above analysis, the number of solutions in total number of encodings (genes) is 𝑛, where “1” denotes
the solution space increases exponentially as 𝑞 increases. construction units, design units, and suppliers that have been
Model (12) is NP-Hard, which implied that model (13) is more selected; “0” indicates construction units, design units, and
difficult to solve. Thus, traditional exhaustion methods are suppliers that have not been selected; blue rectangles denote
only applicable [40] if the problem size is small; namely, 𝑛/ℎ construction units of department 1 and colorless rectangles
and 𝑞 are small at the same time. If the problem size becomes indicate candidate units of department 2. For model (13),
large, intelligent optimization algorithm is preferred. In this the number of genes whose values are 1 is 𝑞. This means
paper, a multiobjective evolutionary algorithm (MOEA) is that 𝑞 units are selected. For instance, if 𝑛 = 6, ℎ = 2,
designed due to larger size of construction units, design units, 𝑞1 = 𝑞2 = 2, 𝑆1 = {𝑝1 , 𝑝2 , 𝑝3 }, 𝑆2 = {𝑝4 , 𝑝5 , 𝑝6 , 𝑝7 },
and suppliers. then there are seven candidate units for coding. There are
Discrete Dynamics in Nature and Society 7

1 1 1 1 Initial population

0 1 1 0 1 0 1 Chromosome

Figure 5: Schematic of encoding rules.

Parent 1

1 0 0 1 0 1 1 0 0 1 0 1 1 0 0 0 1 0

Parent 2

1 0 1 1 0 0 1 0 1 0 0 1 0 0 1 0 0 1

Crossover

Descendant 1

1 0 1 1 0 1 1 0 0 1 0 1 1 0 0 0 1 0

Descendant 2

1 0 0 1 0 0 1 0 1 0 0 1 0 0 1 0 0 1

Figure 6: Pick-and-replace crossover operators.

four units selected, namely, 𝑝2 , 𝑝3 , 𝑝5 , 𝑝7 . Moreover, 𝑝2 and by mating is a major link in genetic and biological evolution
𝑝3 come from department 1 (e.g., design department). 𝑝5 process. Crossover operators are used through imitating
and 𝑝7 come from department 2 (e.g., supplier department). this link and producing new individuals (contractors) in
Encoding rules are shown in Figure 5, where candidate units evolutionary algorithms. So-called crossover operations in
of the design department are 𝑝1 , 𝑝2 , 𝑝3 ; candidate units of evolutionary algorithms refer to two pairs of chromosomes
the supplier department are 𝑝4 , 𝑝5 , 𝑝6 , 𝑝7 . All candidate units which are to exchange some of their genes in some manners
from left to right are 𝑝1 , 𝑝2 , 𝑝3 , 𝑝4 , 𝑝5 , 𝑝6 , 𝑝7 , which constitute and to form two new individuals. The new individuals
a chromosome (all design units and suppliers of bidding). obtain genetic information from two parental individuals
According to the encoding rules, we define values of 𝑛 and 𝑞 in and generate more excellent individuals (selected contractor)
advance. Then we randomly generate the feasible design units than parental individuals randomly.
and suppliers, and an initial population is constituted finally. The design and implementation of crossover operators
We have to explain that tournament selection strategy is are closely related to our proposed problem. In general,
adopted in the algorithm. The specific process is as follows: we require it not to damage fine patterns too much with
fine properties in a string of coding of the individuals
(1) 𝑟 individual units (construction units, design units,
(construction units, design units, and suppliers). Moreover,
and suppliers) are selected randomly in a population
we also require it to produce some better and new individuals
(the three departments) each time according to uni-
effectively. Thus, crossover operators are designed to unify
form distribution.
with individuals coding. Crossover operation adopts double
(2) Individual units (construction units, design units, and cross nodes. The cross process is as follows:
suppliers) with the greatest adaptive values will be
selected as individual units to generate descendants. (1) It randomly generates intersection in two parent
(3) Parameter 𝑟 refers to competition scale, and 𝑟 = 2 individuals.
in this paper. Obviously, this option makes individual (2) Then, it exchanges parts between two nodes of parent
units with better adaptive values have more oppor- individuals.
tunities to “survival” (selected). Meanwhile, relative
(3) It produces descendants finally.
values of the fitness are used to be the choice standard.
Thus, it can avoid influence of the super individual However, using the above cross method may generate unfea-
units, phenomena of premature convergence, and sible solutions. For instance, 𝑛 = 18, 𝑞 = 8, ℎ = 2,
stagnation to a certain extent. 𝑞1 = 5, 𝑞2 = 3, 𝑆1 = {𝑝1 , 𝑝2 , 𝑝3 , 𝑝4 , 𝑝5 , 𝑝6 , 𝑝7 , 𝑝8 , 𝑝9 , 𝑝10 },
𝑆2 = {𝑝11 , 𝑝12 , 𝑝13 , 𝑝14 , 𝑝15 , 𝑝16 , 𝑝16 , 𝑝17 , 𝑝18 }, and the process
4.2. Design and Implement of Crossover Operator. In pro- of pick-and-replace crossover operators is shown in Figure 6.
cess of biological evolution, two homologous chromosomes It can be seen that gene fragment 1 of two descendants
restructure by mating, forming a new chromosome and after crossover operation does not satisfy constraint condi-
thereby producing new individuals or species. Restructure tions 𝑞1 = 5; namely, the solution is an unfeasible solution.
8 Discrete Dynamics in Nature and Society

Existing research shows that repaired strategies of unfeasible The specific process of calculating fitness values of individuals
solutions outclass refused strategies and punished strategies in this paper is described as follows: In the target space,
for many combinatorial optimization problems. Thus, this according to objective function vectors, individuals are com-
paper designs the following repair strategies: we choose one pared in a population and divided into multiple controlled
of the descendants, namely, descendant 1 after any cross. At fronts. Obviously, individuals located in the front of the
this time, we have statistics number 𝑞𝑗∗ of the descendants sequence are superior to the others, named Nondominated
that gene values are “1.” If 𝑞𝑗∗ = 𝑞1 , it is shown that the Sorting. Meanwhile, we introduce the concept of crowded
descendant is a feasible individual, and there is no need to degree. Small crowded degree refers to an individual who
repair it; if 𝑞𝑗∗ ≠ 𝑞1 , statistics number of the descendants has has more points in the current population and bigger dense
the following two kinds of circumstances to repair: distribution. Then, we can get fit and unfit quality in division
according to crowded degrees of all individuals. That is, the
(1) 𝑞𝑗∗ > 𝑞1 : randomly generate 𝑞𝑗∗ − 𝑞1 genes which are greater the crowded degree, the more optimal the individual.
selected randomly from the genes with value of “0” in The method steps are as follows.
the position of genes and the genes with value of “1”
in fragment 1 of descendant 1; then randomly generate (1) Calculate the Genetic Distances between Individuals. For
𝑞𝑗∗ − 𝑞1 genes which are selected randomly from the any two individuals 𝑖 and 𝑗 (𝑗 ≠ 𝑖) in solution space, put
genes with value of “1” in the position of genes and the their chromosomes to be 𝑥𝑖 = [𝑥𝑖1 , 𝑥𝑖2 , . . . , 𝑥𝑖𝑛 ] and 𝑥𝑗 =
genes with value of “0” in fragment 1 of descendant 2. [𝑥𝑗1 , 𝑥𝑗2 , . . . , 𝑥𝑗𝑛 ]. Then, the genetic distance 𝐷𝑖𝑗 between
them can be obtained by the following formula:
(2) 𝑞𝑗∗ < 𝑞1 : randomly generate 𝑞𝑗∗ − 𝑞𝑗 genes which
𝑛
are selected randomly from the genes with value of 󵄨 󵄨
𝐷𝑖𝑗 = ∑ 󵄨󵄨󵄨󵄨𝑥𝑖𝑚 − 𝑥𝑗𝑚 󵄨󵄨󵄨󵄨 . (16)
“1” in the position of genes and the genes with value
𝑚=1
of “0” in fragment 1 of descendant 1; then randomly
generate 𝑞𝑗∗ − 𝑞𝑗 genes which are selected randomly (2) Compare the Genetic Distances among Individuals. Com-
from the genes with value of “0” in the position of pare genetic distances of one individual with the others; then
genes and the genes with value of “1” in fragment 1 of let the minimum distance min{𝐷𝑖𝑗} be crowded degree of the
descendant 2. The reparation ensures that the number individual.
of the genes with value 1 in the descendant equals Compared with Nondominated Sorting approach, the
𝑞1 and guarantees the individual feasibility to ensure above method does not need to calculate objective function
feasibility of the solution. values of all individuals. Thus, it reduces computational
complexity of the algorithm, removes individuals located in
In the same way, the second section of the genes also densely populated area of the solution space, and keeps diver-
conducts similar cross repaired strategies, which can ensure sity of understanding better. According to the argument in the
that the number of fragment 2 of descendants such that gene above, values of 𝑞 determine the solution space. The solution
values are “1” is 𝑞𝑗 (𝑗 = 1, . . . , ℎ). space will increase when 𝑞 increases. In addition, we could
obtain smaller nondominated fronts and shorten calculating
4.3. Design of Mutational Operator. In genetic and biological and sorting time for distances of crowded degrees. Finally,
evolution process, the cell division to copy the link may we could evaluate all the individuals with advantage relations
produce some copy error caused by some accidental factors. and distances of crowded degrees. Based on the advantages
Then, this can lead to some biological gene mutations and disadvantages relations, we could give corresponding
somehow and thus can produce new chromosomes that show fitness values.
new biological traits. Although possibility of this kind of Based on the above elaboration, we summarize the main
variation is small, it is still not negligible to produce new steps of the algorithm which are as follows.
species. To imitate the variation in the process, we introduce
the mutational operators to produce new individuals (con- Step 1 (generating an initial population). Set up necessary
tractors). parameters including generations of population evolution
Mutation operation is introduced to reduce local con- 𝑚𝑎𝑥𝐺𝑒𝑛, size of the population 𝑝𝑜𝑝𝑆𝑖𝑧𝑒, and probability of
vergence [43]. In addition, individuals use reverse mutation crossover and mutation 𝑝𝑐 , 𝑝𝑚 . Generate initial population
operation in designing algorithms which randomly select two 𝑃(𝑡), 𝑡 = 0. Determine adaptive values of all individuals
points to reverse variation in gene mutation operation and according to fitness assignment method.
to reverse the substring between the two points. Clearly, this
operation is only to change orders of the individual genes Step 2 (selection, crossover, and mutation). Adopt tourna-
without affecting the values of the genes and thus not to ment strategy to select firstly. Generate population 𝑃∗ (𝑡)
produce infeasible individuals. according to given probabilities of crossover and mutation 𝑝𝑐 ,
𝑝𝑚 .
4.4. Calculate Individual Fitness Value. Objective function is Step 3 (combining populations). Combine population 𝑃∗ (𝑡)
usually consistent with fitness function in a single objective and population 𝑃(𝑡), 𝑡 = 0. Generate population 𝑃(𝑡 +
decision model, but fitness assignment must consider more 1) constituted by 𝑝𝑜𝑝𝑆𝑖𝑧𝑒 optimal individuals according to
subobjective functions in the multiobjective decision model. fitness.
Discrete Dynamics in Nature and Society 9

Table 1: Indicators system of project bidding.

Indicators Description
Working experience (𝐼1 ) Work years in this field
Qualification rate of completion inspection and acceptance (𝐼2 ) Degree of achieving requirements in the completed projects
Professional knowledge (𝐼3 ) The knowledge related to the engineering task in the field
History of communication (𝐶1 ) The number of times to communicate and study between units
History of cooperation (𝐶2 ) The number of times to help other units for solving problems

Step 4 (judging the algorithm). End the algorithm if 𝑡 + Table 2: Initial information of individual indicators 𝐼1 , 𝐼2 , and 𝐼3 of
1 > 𝑚𝑎𝑥𝐺𝑒𝑛. Put solutions and corresponding vectors of candidate units.
objective function values to be as Pareto optimal solutions Indicators
(approximate) and Pareto optimal (approximate). Otherwise, Candidate units
𝐼1 𝐼2 𝐼3
return to Step 2.
𝑝1 21 5 9
𝑝2 14 1 1
5. Calculation Example 𝑝3 11 3 8
Taking a major bridge project practice in China as problem 𝑝4 30 1 6
background, this section illustrates contractor selection of 𝑝5 18 1 7
bidding in construction projects using the above proposed 𝑝6 20 7 7
model and method. 𝑝7 25 3 4
In order to form a cross-functional contractor, we will 𝑝8 23 4 6
select three candidate units from 12 candidate units which 𝑝9 19 5 5
are distributed in three different functional-departments. The 𝑝10 28 10 6
three departments are design units 𝑆1 , construction units 𝑆2 , 𝑝11 16 4 3
and suppliers 𝑆3 , respectively. And the distribution of the 𝑝12 13 5 1
candidate units 𝑃1 , 𝑃2 , . . . , 𝑃12 in the three departments is as
follows:
𝑆1 = {𝑃1 , 𝑃2 , 𝑃3 } ,
Standardize initial information of individual indicators
𝑆2 = {𝑃4 , 𝑃5 , 𝑃6 , 𝑃7 } , (17) of candidate units using formulas (1), (3), and (4) and
𝑆3 = {𝑃8 , 𝑃9 , 𝑃10 , 𝑃11 , 𝑃12 } . standardized initial information as shown in Table 5.
Standardize initial information of collaborative indicators
In formed cross-functional contractor, it is required that 𝐶1 and 𝐶2 of candidate units using formulas (7), (9), and (10)
one unit is selected from each department, namely, 𝑞1 = and standardized initial information as shown in Tables 6 and
𝑞2 = 𝑞3 = 1. We adopt three individual indicators (working 7.
experience, qualification rate of completion inspection and Assume that weights of individual indicators and collabo-
acceptance, and professional knowledge) and two coordi- rative indicators given directly by DMs are 𝑉 = (0.3, 0.4, 0.3)𝑇
nated indicators (history of communication and history of and 𝑊 = (0.6, 0.4)𝑇 , respectively. We can obtain compre-
cooperation) to select the unit and description as shown in hensive values of individual information and collaborative
Table 1, where 𝐼1 and 𝐶2 are objective indicators; 𝐼2 , 𝐼3 , and 𝐶1 information of candidate units using formulas (5) and (11),
are subjective indicators. On the one hand, values of indicator respectively, as shown in Table 8. Here elements of main
𝐼1 can be obtained through organization records of human diagonal are comprehensive values of individual performance
resources departments. And values of indicator 𝐶2 can be 𝛼𝑖 (𝑖 = 1, . . . , 𝑛), and the other elements are comprehensive
achieved through records of collaborative work systems. On values of collaborative performance 𝛽𝑖𝑗 (𝑖, 𝑗 = 1, . . . , 𝑛; 𝑖 ≠
the other hand, values of indicators 𝐼2 , 𝐼3 , and 𝐶1 can be 𝑗).
scored by experts (1: very poor, 10: very good). The resulting Based on model (13), we obtain the selection model of
initial individual information and collaborative information design units, construction units, and suppliers in construc-
of candidate units are shown in Tables 2–4. tion contractors as follows:

𝑍 = 𝜆 Max 𝑍1 + 𝜇 Max 𝑍2
= (0.64𝑥1 + 0.05𝑥2 + ⋅ ⋅ ⋅ + 0.29𝑥11 + 0.21𝑥12 )
+ [(0.11𝑥1 𝑥2 + 0.46𝑥1 𝑥3 + ⋅ ⋅ ⋅ + 0.25𝑥1 𝑥12 ) + ⋅ ⋅ ⋅ + (0.20𝑥12 𝑥1 + 0.18𝑥12 𝑥2 + ⋅ ⋅ ⋅ + 0.36𝑥12 𝑥11 )] ,
10 Discrete Dynamics in Nature and Society

12

11

Suppliers
10

8
7
3
Co 6 2.5
nst
ru c tio 5 2
nu 1.5 n units
nits 4 1 Desig

Figure 7: Case 1: selection schemes of design units, construction units, and suppliers in construction contractors under the circumstance of
individual performance and collaborative performance of the candidate units have the same weight, namely, 𝜆 = 𝜇 = 0.5.

3
s.t. ∑𝑥𝑖 = 1,
𝑖=1
7
∑𝑥𝑖 = 1,
𝑖=4
12
∑𝑥𝑖 = 1,
𝑖=8

𝑥𝑖 ∈ {0, 1} , 𝑖 = 1, 2, . . . , 12.
(18)

Then, there are three different cases discussed. In case trend, comprehensive performance of case 3 is the best in
1, individual performance and collaborative performance of the three cases, case 1 takes second place, and case 2 is
the candidate units have the same weight, namely, 𝜆 = 𝜇 = the worst. For the optimal contractor of the three cases, as
0.5. We can obtain 60 selection schemes of design units, shown in Table 9, comprehensive performance of the optimal
construction units, and suppliers in construction contractors contractor (design unit 1, construction unit 7, and supplier 11)
as shown in Figure 7, according to descending comprehensive is 𝑍case1 = 2.345 in case 1. Comprehensive performance of
performance. In case 2, individual performance has greater the optimal contractor (design unit 1, construction unit 6, and
weight than collaborative performance. Without loss of gen- supplier 10) is 𝑍case2 = 2.118 in case 2. Comprehensive perfor-
erality, we assume 𝜆 = 0.8, 𝜇 = 0.2. We can obtain 60 mance of the optimal contractor (design unit 1, construction
selection schemes of design units, construction units, and unit 4, and supplier 8) is 𝑍case3 = 2.942 in case 3. Obviously,
suppliers in construction contractors as shown in Figure 8, comprehensive performance in case 3 is the greatest of all the
according to descending comprehensive performance. In case cases. That is to say, DMs could obtain the optimal contractor
3, collaborative performance has greater weight than individ- if they pay more attention to collaborative performance.
ual performance. Without loss of generality, we assumed 𝜆 =
0.2, 𝜇 = 0.8. We can obtain 60 selection schemes of design 6. Conclusions
units, construction units, and suppliers in construction
contractors as shown in Figure 9, according to descending This paper developed a support decision-making system
comprehensive performance. In Figures 7–9, a point stands for contractor selection of bidding in construction projects
for a contractor. Moreover, a point includes one design unit, based on individual indicators and collaborative indicators.
one construction unit, and one supplier denoted by three Construction units, design units, and suppliers are used to
coordinates, respectively. Obviously, these sets of selection form a cross-functional contractor. Then, a cross-functional
schemes of the three cases are different. contractor with 28 candidate units distributed in the three
In addition, we compare comprehensive performance of departments (construction units, design units, and suppliers)
the three cases, as shown in Figure 10. From the overall is used as an example to explain the method. The best
Discrete Dynamics in Nature and Society 11

Table 3: Initial information of collaborative indicator 𝐶1 of candidate units.

𝑝1 𝑝2 𝑝3 𝑝4 𝑝5 𝑝6 𝑝7 𝑝8 𝑝9 𝑝10 𝑝11 𝑝12
𝑝1 — 2 6 7 6 3 7 4 8 7 8 3
𝑝2 2 — 6 2 6 7 3 6 1 2 6 3
𝑝3 6 6 — 6 3 3 1 6 2 6 7 2
𝑝4 7 2 6 — 7 1 2 5 6 8 3 6
𝑝5 6 6 3 7 — 5 6 7 2 4 4 4
𝑝6 3 7 3 1 5 — 4 7 4 5 4 3
𝑝7 7 3 1 2 6 4 — 2 1 7 10 5
𝑝8 4 6 6 5 7 7 2 — 8 3 6 2
𝑝9 8 1 2 6 2 4 1 8 — 3 1 6
𝑝10 7 2 6 8 4 5 7 3 3 — 3 7
𝑝11 8 6 7 3 4 4 10 6 1 3 — 5
𝑝12 3 3 2 6 4 3 5 2 6 7 5 —

Table 4: Initial information of collaborative indicator 𝐶2 of candidate units.

𝑝1 𝑝2 𝑝3 𝑝4 𝑝5 𝑝6 𝑝7 𝑝8 𝑝9 𝑝10 𝑝11 𝑝12

𝑝1 — 4 13 3 1 7 0 5 4 8 9 12
𝑝2 25 — 5 10 5 3 5 9 3 14 3 14
𝑝3 1 15 — 5 10 4 0 4 19 14 17 6
𝑝4 1 7 2 — 13 25 0 2 13 6 13 7
𝑝5 2 9 12 3 — 6 41 0 12 10 3 11
𝑝6 13 13 3 11 10 — 0 14 6 8 13 0
𝑝7 3 0 14 8 12 12 — 8 7 12 9 14
𝑝8 5 5 9 0 9 6 0 — 15 3 6 14
𝑝9 0 13 4 4 3 4 9 7 — 3 0 9
𝑝10 0 8 4 6 6 11 7 5 2 — 11 0
𝑝11 14 8 5 10 10 3 6 4 6 3 — 13
𝑝12 7 5 8 11 8 3 3 15 0 8 9 —

Table 5: Standardized initial information of individual indicators of

candidate units.
12 Indicators
Candidate units
𝐼1 𝐼2 𝐼3
11
𝑝1 0.53 0.44 1.00
Suppliers

10 𝑝2 0.16 0.00 0.00

𝑝3 0.00 0.22 0.88
9 𝑝4 1.00 0.00 0.63
8
𝑝5 0.37 0.00 0.75
7 𝑝6 0.47 0.67 0.75
Co 6
3 𝑝7 0.74 0.22 0.38
nst 2.5
ru c tio 5 2 𝑝8 0.63 0.33 0.63
nu
nits 1.5 gn units 𝑝9 0.42 0.44 0.50
4 1 D es i
𝑝10 0.89 1.00 0.63
Figure 8: Case 2: selection schemes of design units, construc- 𝑝11 0.26 0.33 0.25
tion units, and suppliers in construction contractors under the
circumstance of individual performance have greater weight than 𝑝12 0.11 0.44 0.00
collaborative performance. Without loss of generality, we assumed
𝜆 = 0.8, 𝜇 = 0.2.
12 candidate units which are distributed in three different
functional-departments (construction units, design units,
individual performance and collaborative performance are and suppliers) is used to illustrate our method. As a result of
used to formulate the problem as a binary optimization limitation of resources and different preference of DMs, we
problem. An example including three candidate units from can further modify the model and increase some objectives
12 Discrete Dynamics in Nature and Society

Table 6: Standardized initial information of collaborative indicators 𝐶1 of candidate units.

𝑝1 𝑝2 𝑝3 𝑝4 𝑝5 𝑝6 𝑝7 𝑝8 𝑝9 𝑝10 𝑝11 𝑝12

𝑝1 — 0.11 0.56 0.67 0.56 0.22 0.67 0.33 0.78 0.67 0.78 0.22
𝑝2 0.11 — 0.56 0.11 0.56 0.67 0.22 0.56 0.00 0.11 0.56 0.22
𝑝3 0.56 0.56 — 0.56 0.22 0.22 0.00 0.56 0.11 0.56 0.67 0.11
𝑝4 0.67 0.11 0.56 — 0.67 0.00 0.11 0.45 0.56 0.78 0.22 0.56
𝑝5 0.56 0.56 0.22 0.67 — 0.45 0.56 0.67 0.11 0.33 0.33 0.33
𝑝6 0.22 0.67 0.22 0.00 0.45 — 0.33 0.67 0.33 0.45 0.33 0.22
𝑝7 0.67 0.22 0.00 0.11 0.56 0.33 — 0.11 0.00 0.67 1.00 0.45
𝑝8 0.33 0.56 0.56 0.45 0.67 0.67 0.11 — 0.78 0.22 0.56 0.11
𝑝9 0.78 0.00 0.11 0.56 0.11 0.33 0.00 0.78 — 0.22 0.00 0.56
𝑝10 0.67 0.11 0.56 0.78 0.33 0.45 0.67 0.22 0.22 — 0.22 0.67
𝑝11 0.78 0.56 0.67 0.22 0.33 0.33 1.00 0.56 0.00 0.22 — 0.45
𝑝12 0.22 0.22 0.11 0.56 0.33 0.22 0.45 0.11 0.56 0.67 0.45 —

Table 7: Standardized initial information of collaborative indicators 𝐶2 of candidate units.

𝑝1 𝑝2 𝑝3 𝑝4 𝑝5 𝑝6 𝑝7 𝑝8 𝑝9 𝑝10 𝑝11 𝑝12
𝑝1 — 0.10 0.32 0.07 0.02 0.17 0.00 0.12 0.10 0.20 0.22 0.29
𝑝2 0.61 — 0.12 0.24 0.12 0.07 0.12 0.22 0.07 0.34 0.07 0.34
𝑝3 0.02 0.37 — 0.12 0.24 0.10 0.00 0.10 0.46 0.34 0.41 0.15
𝑝4 0.02 0.17 0.05 — 0.32 0.61 0.00 0.05 0.32 0.15 0.32 0.17
𝑝5 0.05 0.22 0.29 0.07 — 0.15 1.00 0.00 0.29 0.24 0.07 0.27
𝑝6 0.32 0.32 0.07 0.27 0.24 — 0.00 0.34 0.15 0.20 0.32 0.00
𝑝7 0.07 0.00 0.34 0.20 0.29 0.29 — 0.20 0.17 0.29 0.22 0.34
𝑝8 0.12 0.12 0.22 0.00 0.22 0.15 0.00 — 0.37 0.07 0.15 0.34
𝑝9 0.00 0.32 0.10 0.10 0.07 0.10 0.22 0.17 — 0.07 0.00 0.22
𝑝10 0.00 0.20 0.10 0.15 0.15 0.27 0.17 0.12 0.05 — 0.27 0.00
𝑝11 0.34 0.20 0.12 0.24 0.24 0.07 0.15 0.10 0.15 0.07 — 0.32
𝑝12 0.17 0.12 0.20 0.27 0.20 0.07 0.07 0.37 0.00 0.20 0.22 —

Table 8: Comprehensive values of individual information and collaborative information of candidate units.
𝑝1 𝑝2 𝑝3 𝑝4 𝑝5 𝑝6 𝑝7 𝑝8 𝑝9 𝑝10 𝑝11 𝑝12
𝑝1 0.64 0.11 0.46 0.43 0.34 0.20 0.40 0.25 0.51 0.48 0.56 0.25
𝑝2 0.31 0.05 0.38 0.16 0.38 0.43 0.18 0.42 0.03 0.20 0.36 0.27
𝑝3 0.34 0.48 0.35 0.38 0.23 0.17 0.00 0.38 0.25 0.47 0.57 0.13
𝑝4 0.41 0.13 0.36 0.19 0.53 0.24 0.07 0.29 0.46 0.53 0.28 0.40
𝑝5 0.36 0.42 0.25 0.43 0.34 0.33 0.74 0.40 0.18 0.29 0.23 0.31
𝑝6 0.26 0.53 0.16 0.11 0.37 0.63 0.20 0.54 0.26 0.35 0.33 0.13
𝑝7 0.43 0.13 0.14 0.15 0.45 0.31 0.42 0.15 0.07 0.52 0.69 0.41
𝑝8 0.25 0.38 0.42 0.27 0.49 0.46 0.07 0.51 0.62 0.16 0.40 0.20
𝑝9 0.47 0.13 0.11 0.38 0.09 0.24 0.09 0.54 0.45 0.16 0.00 0.42
𝑝10 0.40 0.15 0.38 0.53 0.26 0.38 0.47 0.18 0.15 0.86 0.24 0.40
𝑝11 0.60 0.42 0.45 0.23 0.29 0.23 0.66 0.38 0.06 0.16 0.29 0.40
𝑝12 0.20 0.18 0.15 0.44 0.28 0.16 0.30 0.21 0.34 0.48 0.36 0.21

Table 9: Comprehensive performance of the three cases.

Individual Collaborative Comprehensive
Candidate units
Cases performance performance performance
Design units Construction units Suppliers 𝑍1 𝑍2 𝑍
Case 1 1 7 11 1.35 3.34 2.345
Case 2 1 6 10 2.13 2.07 2.118
Case 3 1 4 8 1.34 1.9 2.942
Discrete Dynamics in Nature and Society 13

𝑔: Number of individual indicators

𝑘: Number of collaborative indicators
12 𝑝𝑖 : Candidate unit 𝑖, 𝑖 = 1, 2, . . . , 𝑛
𝑆𝑗 : Set of candidate units in department 𝑗,
11 𝑗 = 1, . . . , ℎ
𝑁𝑗 : Indicator set of candidate units in set 𝑆𝑗 ,
Suppliers

10 𝑗 = 1, . . . , ℎ
𝐼𝑙 : Individual indicator 𝑙, 𝑙 = 1, . . . , 𝑔
9 𝐶𝑚 : Collaborative indicator 𝑚, 𝑚 = 1, . . . , 𝑘
𝑔
V𝑙 : Weight of individual indicator 𝐼𝑙 , ∑𝑙=1 V𝑙 = 1;
8
7 0 ≤ V𝑙 ≤ 1
6 3 𝑤𝑚 : Weight of collaborative indicator 𝐶𝑚 ,
Con 2.5
stru
ctio 5 2 ∑𝑘𝑚=1 𝑤𝑚 = 1; 0 ≤ 𝑤𝑚 ≤ 1
nu
nits 1.5 n units 𝑥𝑖 : Decision variables, 𝑥𝑖 = 1 denotes that unit
4 1 Desig
𝑝𝑖 is selected; or else, 𝑥𝑖 = 0.
Figure 9: Case 3: selection schemes of design units, construc-
tion units, and suppliers in construction contractors under the
circumstance of collaborative performance have greater weight than Competing Interests
individual performance. Without loss of generality, we assumed 𝜆 =
The authors declare that they have no competing interests
0.2, 𝜇 = 0.8.
regarding the publication of this paper.
3
Acknowledgments
2.5 This work was partly supported by the National Natu-
Comprehensive performance

ral Science Foundation of Key Projects (no. 71390520),

National Natural Science Foundation of China (nos. 71271103,
2
71101067, and 71571098), and Nanjing University Graduate
Students Foundation of Scientific Innovation Projects (no.
1.5 2014CW05).

References
1
[1] R. Awwad, “Evolutionary simulation of contractors’ learning
and behavior under two bid-tendering approaches,” Journal of
0.5 Management in Engineering, vol. 32, no. 2, Article ID 04015041,
0 10 20 30 40 50 60
2016.
Feasible solution sequence
[2] S. Asgari, R. Awwad, A. Kandil, and I. Odeh, “Impact of
Case 1 considering need for work and risk on performance of con-
Case 2 struction contractors: an agent-based approach,” Automation in
Case 3 Construction, vol. 65, pp. 9–20, 2016.
[3] J.-S. Chou, “Generalized linear model-based expert system for
Figure 10: Comprehensive performance of selection schemes of estimating the cost of transportation projects,” Expert Systems
design units, construction units, and suppliers in construction with Applications, vol. 36, no. 3, pp. 4253–4267, 2009.
contractors according to the above three cases.
[4] M.-Y. Cheng, H.-C. Tsai, and E. Sudjono, “Conceptual cost
estimates using evolutionary fuzzy hybrid neural network for
projects in construction industry,” Expert Systems with Applica-
and constraints for the model. In addition, theory thoughts tions, vol. 37, no. 6, pp. 4224–4231, 2010.
put forward in this paper are hoping to bring some enlight- [5] D. K. H. Chua and D. Li, “Key factors in bid reasoning model,”
enment for bidding problems in construction projects. Journal of Construction Engineering and Management, vol. 126,
no. 5, pp. 349–357, 2000.
[6] P. Ballesteros-Pérez, M. L. del Campo-Hitschfeld, D. Mora-
Notation Melià, and D. Domı́nguez, “Modeling bidding competitiveness
𝑛: Total number of candidate units and position performance in multi-attribute construction auc-
ℎ: The total number of departments tions,” Operations Research Perspectives, vol. 2, pp. 24–35, 2015.
𝑛𝑗 : Number of candidate units in department [7] Z. Jin, F. Deng, H. Li, and M. Skitmore, “Practical framework for
𝑗, ∑ℎ𝑗=1 𝑛𝑗 = 𝑛 measuring performance of international construction firms,”
Journal of Construction Engineering and Management, vol. 139,
𝑞: Sum of departments no. 9, pp. 1154–1167, 2013.
𝑞𝑗 : Number of selected units in department 𝑗, [8] T.-H. Sohn, H.-R. Kim, and H.-S. Jang, “An MDB business
∑ℎ𝑗=1 𝑞𝑗 = 𝑞 competency assessment of Korean construction companies,”
14 Discrete Dynamics in Nature and Society

KSCE Journal of Civil Engineering, vol. 18, no. 5, pp. 1314–1321, [25] Y. I. Topcu, “A decision model proposal for construction
2014. contractor selection in Turkey,” Building and Environment, vol.
[9] A. Mahdavi and M. Hastak, “Quantitative analysis of bidding 39, no. 4, pp. 469–481, 2004.
strategies: a hybrid agent based-system dynamics approach,” in [26] B. McCabe, V. Tran, and J. Ramani, “Construction prequalifica-
Proceedings of the Construction Research Congress: Construction tion using data envelopment analysis,” Canadian Journal of Civil
in a Global Network (CRC ’14), pp. 1129–1138, May 2014. Engineering, vol. 32, no. 1, pp. 183–193, 2005.
[10] S. D. Jap and P. A. Naik, “BidAnalyzer: a method for estimation [27] K. K. Lai, S. L. Liu, and S. Y. Wang, “A method used for evalu-
and selection of dynamic bidding models,” Marketing Science, ating bids in the Chinese construction industry,” International
vol. 27, no. 6, pp. 949–960, 2008. Journal of Project Management, vol. 22, no. 3, pp. 193–201, 2004.
[11] I. Yu, K. Kim, Y. Jung, and S. Chin, “Comparable performance [28] E. W. L. Cheng and H. Li, “Contractor selection using the
measurement system for construction companies,” Journal of analytic network process,” Construction Management and Eco-
Management in Engineering, vol. 23, no. 3, pp. 131–139, 2007. nomics, vol. 22, no. 10, pp. 1021–1032, 2004.
[12] C. M. Harper, K. R. Molenaar, S. Anderson, and C. Schex- [29] H. Missbauer and W. Hauber, “Bid calculation for construction
nayder, “Synthesis of performance measures for highway cost projects: regulations and incentive effects of unit price con-
estimating,” Journal of Management in Engineering, vol. 30, no. tracts,” The European Journal of Operational Research, vol. 171,
3, Article ID 04014005, 2014. no. 3, pp. 1005–1019, 2006.
[13] D. Schiereck and J. Vogt, “Long-run M & A success of strategic [30] S. Lambropoulos, “The use of time and cost utility for con-
bidders in the construction industry,” Problems and Perspectives struction contract award under European Union Legislation,”
in Management, vol. 11, no. 4, pp. 46–67, 2013. Building and Environment, vol. 42, no. 1, pp. 452–463, 2007.
[14] M. Wanous, A. H. Boussabaine, and J. Lewis, “To bid or not [31] S. S. Padhi and P. K. J. Mohapatra, “Centralized bid evaluation
to bid: a parametric solution,” Construction Management and for awarding of construction projects—a case of India govern-
Economics, vol. 18, no. 4, pp. 457–466, 2000. ment,” International Journal of Project Management, vol. 28, no.
[15] A. Bagies and C. Fortune, “Bid/ no-bid decision modelling for 3, pp. 275–284, 2010.
construction projects,” in Proceedings of the 22nd Annual Con- [32] B. S. Liu, T. F. Huo, P. C. Liao, J. Gong, and B. Xue, “A group
ference on Association of Researchers in Construction Manage- decision-making aggregation model for contractor selection in
ment (ARCOM ’06), pp. 511–521, Birmingham, UK, September large scale construction projects based on two-stage partial least
2006. squares (PLS) path modeling,” Group Decision & Negotiation,
[16] M. S. El-Mashaleh, “Empirical framework for making the vol. 24, no. 5, pp. 855–883, 2014.
Bid/No-Bid decision,” Journal of Management in Engineering, [33] A. T. de Almeida, “Multicriteria decision model for outsourcing
vol. 29, no. 3, pp. 200–205, 2013. contracts selection based on utility function and ELECTRE
[17] M.-Y. Cheng, C.-C. Hsiang, H.-C. Tsai, and H.-L. Do, “Bidding method,” Computers & Operations Research, vol. 34, no. 12, pp.
decision making for construction company using a multi- 3569–3574, 2007.
criteria prospect model,” Journal of Civil Engineering and [34] L. H. Alencar and A. T. de Almeida, “Multicriteria decision
Management, vol. 17, no. 3, pp. 424–436, 2011. group model for the selection of suppliers,” Pesquisa Opera-
[18] I. Dikmen, M. T. Birgonul, and A. K. Gur, “A case-based deci- cional, vol. 28, no. 2, pp. 321–337, 2008.
sion support tool for bid mark-up estimation of international [35] A. J. de Melo Brito, A. de Almeida Filho, and A. T. de
construction projects,” Automation in Construction, vol. 17, no. Almeida, “Multi-criteria decision model for selecting repair
1, pp. 30–44, 2007. contracts by applying utility theory and variable interdependent
[19] J.-S. Chou, A.-D. Pham, and H. Wang, “Bidding strategy parameters,” IMA Journal of Management Mathematics, vol. 21,
to support decision-making by integrating fuzzy AHP and no. 4, pp. 349–361, 2010.
regression-based simulation,” Automation in Construction, vol. [36] F. Chiclana, E. Herrera-Viedma, F. Herrera, and S. Alonso,
35, pp. 517–527, 2013. “Some induced ordered weighted averaging operators and their
[20] D. J. Lowe and J. Parvar, “A logistic regression approach use for solving group decision-making problems based on
to modelling the contractor’s decision to bid,” Construction fuzzy preference relations,” The European Journal of Operational
Management and Economics, vol. 22, no. 6, pp. 643–653, 2004. Research, vol. 182, no. 1, pp. 383–399, 2007.
[21] B. Liu, T. Huo, J. Meng, J. Gong, Q. Shen, and T. Sun, [37] I. J. Pérez, F. J. Cabrerizo, S. Alonso, and E. Herrera-Viedma,
“Identification of key contractor characteristic factors that “A new consensus model for group decision making problems
affect project success under different project delivery systems: with non-homogeneous experts,” IEEE Transactions on Systems,
empirical analysis based on a group of data from China,” Journal Man, and Cybernetics: Systems, vol. 44, no. 4, pp. 494–498, 2014.
of Management in Engineering, vol. 32, no. 1, Article ID 5015003, [38] A. Nosratinia and T. E. Hunter, “Grouping and partner selection
2016. in cooperative wireless networks,” IEEE Journal on Selected
[22] E. A. Demirtas and Ö. Üstün, “An integrated multiobjective Areas in Communications, vol. 25, no. 2, pp. 369–378, 2007.
decision making process for supplier selection and order allo- [39] C. C. Kuo, F. Glover, and K. S. Dhir, “Analyzing and modeling
cation,” Omega, vol. 36, no. 1, pp. 76–90, 2008. the maximum diversity problem by Zero-one programming,”
[23] F. Khosrowshahi, “Neural network model for contractors’ pre- Decision Sciences, vol. 24, no. 6, pp. 1171–1185, 1993.
qualification for local authority projects,” Engineering, Con- [40] R. K. Gupta, K. S. Shishodia, and G. S. Sekhon, “Optimization
struction and Architectural Management, vol. 6, no. 3, pp. 315– of grinding process parameters using enumeration method,”
328, 1999. Journal of Materials Processing Technology, vol. 112, no. 1, pp. 63–
[24] I. M. Mahdi, M. J. Riley, S. M. Fereig, and A. P. Alex, “A 67, 2001.
multi-criteria approach to contractor selection,” Engineering [41] Q. Long, C. Wu, T. Huang, and X. Wang, “A genetic algorithm
Construction and Architectural Management, vol. 9, no. 1, pp. for unconstrained multi-objective optimization,” Swarm and
29–37, 2002. Evolutionary Computation, vol. 22, pp. 1–14, 2015.
Discrete Dynamics in Nature and Society 15

[42] Q. Long and C. Wu, “A hybrid method combining genetic

algorithm and Hooke-Jeeves method for constrained global
optimization,” Journal of Industrial and Management Optimiza-
tion, vol. 10, no. 4, pp. 1279–1296, 2014.
[43] H. Holland, Adaptation in Natural and Artificial Systems: An
Introductory Analysis with Applications to Biology, Control, and
Artificial Intelligence, University of Michigan Press, Ann Arbor,
Mich, USA, 1992.
Advances in Advances in Journal of Journal of
Operations Research
Hindawi Publishing Corporation
Decision Sciences
Hindawi Publishing Corporation
Applied Mathematics
Hindawi Publishing Corporation
Algebra
Hindawi Publishing Corporation
Probability and Statistics
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

The Scientific International Journal of

World Journal
Hindawi Publishing Corporation
Differential Equations
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Submit your manuscripts at

http://www.hindawi.com

International Journal of Advances in

Combinatorics
Hindawi Publishing Corporation
Mathematical Physics
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Journal of Journal of Mathematical Problems Abstract and Discrete Dynamics in

Complex Analysis
Hindawi Publishing Corporation
Mathematics
Hindawi Publishing Corporation
in Engineering
Hindawi Publishing Corporation
Applied Analysis
Hindawi Publishing Corporation
Nature and Society
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

International
Journal of Journal of
Mathematics and
Mathematical
Discrete Mathematics
Sciences

Journal of International Journal of Journal of

Hindawi Publishing Corporation Hindawi Publishing Corporation Volume 2014

Function Spaces
Hindawi Publishing Corporation
Stochastic Analysis
Hindawi Publishing Corporation
Optimization
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014