Multimodal optimization for time-cost trade-off in construction projects

using a novel hybrid method based on FA and PSO

Optimización multimodal para la compensación de tiempo-costo, en proyectos de construcción,
usando un nuevo método híbrido, basado en FA y PSO

Gülçağ Albayrak (Main and Corresponding Author)

Dept. of Civil Engineering, Eskisehir Osmangazi University
ESOGU Meselik Campus, 26480, Eskisehir, (Turkey)
[email protected]

İlker Özdemir
Dept. of Civil Engineering, Eskisehir Osmangazi University
ESOGU Meselik Campus, 26480, Eskisehir, (Turkey)
[email protected]

Manuscript Code: 1051

Date of Acceptance/Reception: 01.08.2018/29.12.2017
DOI: 10.7764/RDLC.17.2.304

Abstract
Completion of the activities within optimal time and cost plays a significant role in construction projects. Recently, project managers have to decrease
the total durations and costs of the projects more than before due to the competitive environment. Mostly, decision makers usually seek different
alternatives which reduce time and cost. As well as being one of the most major topics of construction management, this problem called time-cost
trade-off (TCTO) which is extremely difficult to solve with traditional mathematical methods. In recent years, metaheuristic algorithms are outstanding
methods in this field due to their flexible and adaptable structure. This paper presents a new algorithm called F-PSO which consists of hybridizing
Firefly Algorithm (FA) with Particle Swarm Optimization (PSO). In this method, the problem is modelled with various execution modes to select the
optimal one for each activity. The applicability and validity of the proposed method is confirmed by performing 18-activity project as a benchmark
problem. Comparison of numerical results with different metaheuristic algorithms demonstrates the effectiveness and efficiency of F-PSO with regard
to optimality of time and cost outcomes.

Key words: Hybrid algorithm, metaheuristic method, project planning, time-cost trade-off.

Resumen
La finalización de las actividades en un tiempo y costo óptimos desempeña un papel importante en los proyectos de construcción. Recientemente,
los gerentes de proyecto tienen que disminuir la duración total y los costos de los proyectos más que antes debido al entorno competitivo. En general,
quienes toman las decisiones generalmente buscan diferentes alternativas que reduzcan el tiempo y el costo. Además de ser uno de los temas más
importantes de la gestión de la construcción, este problema se llama compensación - tiempo-costo (TCTO), que es extremadamente difícil de resolver
con métodos matemáticos tradicionales. En los últimos años, los algoritmos metaheurísticos son métodos sobresalientes en este campo debido a su
estructura flexible y adaptable. Este artículo presenta un nuevo algoritmo llamado F-PSO que consiste en hibridar el Algoritmo Firefly (FA) con la
Optimización de Enjambre de Partículas (PSO). En este método, el problema se modela con varios modos de ejecución para seleccionar el óptimo
para cada actividad. La aplicabilidad y validez del método propuesto se confirma realizando un proyecto de 18 actividades como un problema de
referencia. La comparación de resultados numéricos con diferentes algoritmos metaheurísticos demuestra la efectividad y eficiencia de F-PSO con
respecto a la optimización de los resultados de tiempo y costo.

Palabras clave: Algoritmo híbrido, método metaheurístico, planificación de proyectos, compensación de tiempo-costo.

Introduction

Generally, all activities are considered with normal durations in traditional project planning (Klerides &
Hadjiconstantinou, 2015). However, the decision makers sometimes have to complete the project earlier due to various
reasons such as unforeseen delays, incentive contracts, imposed deadlines, contract commitments, overhead costs and
pressure to move resources to other project. For this purpose, some of the activities must be accelerated to reduce the
total project duration. This acceleration called crashing which can be achieved by additional resources or new
technological changes, but they always cause increases in cost. Undoubtedly, the duration has a strong relation with
the resource and they act each other inversely proportional. On the other hand, different time-cost relationships can
occur depending on the cost components of the project. The most traditional form of time-cost relationship is expressed
as linear. Linear time-cost relationship is a conventional deterministic model for defining the time-cost problems and
numerous studies have been presented in the literature formerly. Kelley (1961), Fulkerson (1961), Siemens (1971),
Salem & Elmaghraby (1994), Hendrickson & Au (1989), Tareghian & Taheri (2006), and Chen & Tsai (2011) put forward
linear approaches for time-cost problems.

304
The common point of these studies was that the relationship between time and cost of an activity was assumed as a
linear function and the main purpose was to schedule the activities for minimizing the project cost, at that time
(Vanhoucke & Debels, 2007). Despite the fact that the linear approach is simply applicable as an exact method, it does
not reflect the real-life problems precisely enough. Also exact methods can only optimize single-objective problems.
However, if the optimization of time and cost is requested simultaneously, these methods fail. This optimization
problem is known as time-cost trade-off (TCTO) problem in the literature (Hegazy, 1999). The concept of TCTO has been
raised with the multi execution modes which include different time-cost alternatives for activities. This type of TCTO
models called discrete optimization has combination of discrete points corresponding to each time-cost set of the
activities. In project planning, the activities which have at least two modes are considered as multimodal optimization
problems. The aim of TCTO is to find the best alternative solution by providing the optimal total duration and cost of
the project.

Our intent with this paper is to present the applicability of an alternative hybrid intelligent search method for solving
the problem of the minimization of total project cost and duration, when discrete time-cost combinations are available
on the activities of a project. This paper is different from the previous studies with respect to proposing a novel method
for the solution. Moreover, presenting a Pareto front of results to decision makers gives them the opportunity to select
the better solution due to project limitations and make the decision making procedure more flexible. The mathematical
model of TCTO is described considering multimodal and multi-objective optimization requirements. Also a novel hybrid
algorithm (Firefly-Particle Swarm Optimization, abbr. F-PSO) is proposed to solve TCTO problem. F-PSO is based on PSO
combining with FA approach for TCTO problems in construction projects. In this algorithm, PSO searches the solution
space globally when FA focuses on finding optimal solutions. Thus, the advantages of both nature inspired algorithms
are combined. Furthermore, the proposed F-PSO method is tested on 18-activity TCTO problem which is one of the most
well-known problems derived from the literature. For demonstrating the performance of F-PSO, the obtained numerical
solutions are compared with other studies conducted with different metaheuristic methods existing in the literature.

The remaining content of this paper is organized as follows: TCTO is introduced and the mathematical model is explained
in the section of Description of the Problem. The related works which apply metaheuristic methods on TCTO are
mentioned in the section of State of Art. In the section of Methodology, Firefly Algorithm (FA) and Particle Swarm
Optimization (PSO) are summarized and F-PSO is presented. Also, application of F-PSO on the case example is given in
the same section. In the section of Results, the computational results are discussed. Finally, conclusions and directions
of future work are mentioned in the section of Conclusions.

Description of the Problem

TCTO is a multi-objective optimization which aims to minimize both time and cost. The optimization brings the optimal
time-cost sets which constitute Pareto front. Pareto front was first formulated by Pareto (1906) as the commonly
accepted tool for evaluating potential solutions. Pareto front is a set of non-dominated solutions being chosen as
optimal if no objective can be improved without sacrificing at least one other objective (Fan, You, & Li, 2013). Thus, the
decision maker can find the most appropriate time-cost set, which is undetectable by conventional methods, according
to subjective preferences.

The mathematical model of TCTO problem given in the below is adapted from a previous study which is conducted by
the first author of this paper (Albayrak, 2017). The mathematical model is described by Equations (1- 4) as constraints
and Eq. (5) and (6) as objective functions. In Equations (1 - 6), ct, tt, cij and xij represent total cost, total duration, cost of
the jth mode for ith activity, assignment of the jth mode for ith activity respectively. T, mn, n, Tij and Tmax represents starting
time, mode alternatives, activity number, duration of jth mode of ith activity and maximum completion time respectively.
According to the Eq. (1), the project starts at time 0. Eq. (2) states that, the sum of the starting time of the last activity
with duration should be less than or equal to the maximum completion time of the project. According to Eq. (3), the
sum of the starting time of a predecessor activity and the duration of jth mode should be less than or equal to starting
time of the successor activity. The precedence constraints must not be violated. The last constraint Eq. (4) expresses
that only one mode must be assigned for each activity. Accordingly, xij is a binary variable which takes the value of 0-1.

Constraint functions:
𝑇1 = 0; (1)
𝑚𝑛
𝑇𝑛 + ∑𝑗=1 𝑇𝑛𝑗 . 𝑥𝑛𝑗 ≤ 𝑇𝑚𝑎𝑥 ; (2)
𝑚𝑎
𝑇𝑎 + ∑𝑗=1 𝑇𝑎𝑗 . 𝑥𝑎𝑗 ≤ 𝑇𝑏 ; 𝑎 → 𝑏 for all predecessors a,b=1,…,n; (3)
𝑚𝑖
∑𝑛𝑖=1 ∑𝑗=1 𝑥𝑖𝑗 = 1; (4)
305
Objective functions:
𝑚𝑖
𝑀𝑖𝑛 𝑐𝑡 = ∑𝑛𝑖=1 ∑𝑗=1 𝑐𝑖𝑗 𝑥𝑖𝑗 (5)
𝑚𝑛
𝑀𝑖𝑛 𝑡𝑡 = [𝑇𝑛 + ∑𝑗=1(𝑇𝑖𝑗 . 𝑥𝑖𝑗 )] (6)

First objective function Eq. (5) aims to minimize total project cost and second objective function Eq. (6) aims to minimize
total project duration.

State of the Art

The time–cost trade-off problem has been extensively studied decades ago under the assumption of continuous time-
cost relationships. On the contrary, the literature for the case where the time-cost modes are defined at discrete points
(representing distinct alternatives) is relatively recent, despite the fact that the discrete time-cost modes are considered
to be a more realistic model of real projects. Undoubtedly, the problem has gained considerable attention. Because
mathematical structure of the problem includes complicated formulation and, thus, it is very difficult to obtain optimal
solutions efficiently. De, Dunne, Ghosh, & Wells (1997) and also Deineko & Woeginger (2001), showed that this problem
belongs to NP-hard (non-deterministic polynomial time) problem. Due to unlikely existence of any polynomial algorithm
to solve this problem optimally, the efforts have turned to finding the approximation and heuristics methods. The
metaheuristic algorithms, which include a number of different methods, have an important role in this field because of
their widespread application area than conventional methods such as critical path method and linear programming.

Li & Love (1997), Feng, Liu, & Burns (1997; 2000), Leu & Yang (1999) and (Hegazy, 1999) are the initiative studies in the
literature in terms of proposing metaheuristic method based on Genetic Algorithm (GA) to solve TCTO. Thereafter TCTO
has been extensively studied and many different algorithms have been suggested by many researchers. New powerful
metaheuristic algorithms, such as Simulated Annealing, Tabu Search, Ant Colony Algorithm (ACO), Particle Swarm
Optimization and Shuffled Frog Leaping Algorithm, have contributed to further development of optimization. The large
number of metaheuristics proposed to solve the problem, as well as the values to be decided for many parameters of a
metaheuristic, impose their evaluation with regard to their time efficiency and the quality of the obtained solutions.
The studies related to TCTO solving with metaheuristic algorithms are as follows mainly: Elbeltagi, Hegazy, & Grierson
(2005) presented five evolutionary-based optimization algorithms with comparison. Tareghian & Taheri (2007)
introduced the electromagnetic scatter search algorithm for discrete TCTO problem. Huang, Deng, & Zhang (2008) and
Ng & Zhang (2008) proposed ACO. Eshtehardian, Afshar, & Abbasnia (2009) presented a fuzzy multi-objective GA
approach for uncertain environment conditions. Geem (2010) applied a Harmony Search (HS) algorithm to solve time-
cost trade-off problems. The objective is to minimize total cost, which consists of direct cost and time related indirect
cost. Mokhtari, Baradaran Kazemzadeh, & Salmasnia (2011) applied the ant colony system to the multimodal discrete
TCTO problem with normal distribution of activity time values. Sonmez & Bettemir (2012) proposed a new hybrid
algorithm using Simulated Annealing, Quantum Based Simulated Annealing and GA. Cheng & Tran (2014) proposed a
novel approach by introducing their two-phase differential evolution model which was able to successfully reflect both
time-cost effects and resource constraints. Prascevic & Prascevic (2014) applied PSO method for solving time-cost
optimization, and developed for that purpose an appropriate computer programme. Lee, Yi, Lee, & Arditi (2015) used
the existing data from the project schedules for each individual task to find optimal set of parameters for GA as
Advanced Stochastic Time-Cost Tradeoff (ASTCT) method to solve TCTO problem. Ashuri & Tavakolan (2015) considered
the Pareto front optimization of resources along with the time and cost and presented a Shuffled Frog Leaping
Algorithm. Pathak & Srivastava (2015) developed a new technique for project planners for TCTO in fuzzy environment.
Kaveh, Khanzadi, Alipour, & Naraky (2015) showed that two new metaheuristic algorithms, Charged System Search (CSS)
and Colliding Body Optimization (CBO), are utilized for solving this problem. The results show that both of these
algorithms find reasonable solutions; however, CBO could find the result in a less computational time having a better
quality. Shen, Hassani, & Shi (2016) considered the problem with Coub-Douglas production function and hybrid GA.
Hou, Zhao, Wu, Moon, & Wang (2017) formulated a FA to target the optimal combination of the project makespan (start
time, finish time) and execution mode of each project activity by using a series of unique mathematical models. Elloumi,
Fortemps, & Loukil (2017) developed a novel evolutionary algorithm for the problem under mode change disruption.
Zheng & Zhong (2017) presented hybrid GA for TCTO problem considering the environmental impact. He, He, Liu, &
Wang (2017) suggested Variable Neighbourhood Search and Tabu Search for TCTO problem to minimize the maximal
cash flow gap.

On the other hand, the researchers who studied multi objective optimization problems have focused on the hybrid
metaheuristic methods. Mehdizadeh, Hajipour, & Mohammadizadeh (2015) investigated a multi-item capacitated lot-
sizing problem. They develop a bi-objective mathematical programming model with two conflicting objectives. They
306
propose two novel Pareto-based multi-objective metaheuristic algorithms called Multi-objective Vibration Damping
Optimization and the Non-dominated Ranking GA. Alaghebandha, Hajipour, & Hemmati (2017) studied on the
formulation of a sequencing problem with the dual goals of varying the parts utilization and varying the workload which
are typically inversely correlated with each other, and therefore the simultaneous optimization of both is challenging.
Owing to the NP-hardness of the problem, they introduce a discrete PSO algorithm, memetic algorithm, Weighted Sum
Multi-objective GA and Non-dominated Sorting GA to solve problem where these objectives are to be optimized
simultaneously.

Methodology
Metaheuristic Methods

The mathematical complexity of TCTO problem is directly affected by the size of the project network. In other words,
the complexity of the problem grows exponentially as the number of the activities. Therefore the problem of TCTO is
strongly NP-hard which means that TCTO cannot be solved with deterministic solution methods (Deineko & Woeginger,
2001).

Recently, TCTO problems have been solved using metaheuristic methods. Metaheuristic methods can be useful and
effective especially when a problem is difficult to solve with exact methods (Sörensen, Sevaux, & Glover, 2018). They
are also preferred frequently when the research field is large and complicated. Also, achieving the approximate results
in a short time provides sufficient accuracy. These algorithms give quite reasonable solutions, but do not guarantee
exactly optimum results (Yang, Deb, & Fong, 2014).

Hybrid optimization technique is a successful combination of metaheuristic algorithm with another optimization
algorithm that can display a more robust behavior and exhibit greater flexibility against complex and difficult problems
(Blum, Aguilera, Roli, & Sampels, 2008). In this study, integration of these two swarm intelligence metaheuristics is
aimed for solving TCTO problem. The metaheuristics, PSO and FA have recently proved to be successful approaches to
solve complex optimization problems using swarm intelligence. PSO is a powerful optimization technique for solving
multimodal discrete optimization problems. Similarly, FA that imitates flashing behavior of fireflies is known to be
efficient and robust for solution of combinatorial optimization problems (Shelokar, Siarry, Jayaraman, & Kulkarni, 2007).

Recently, hybridizing local search strategies with PSO and FA has attracted more attentions, and the experimental
results verify the promising comprehensive performance, including faster convergence speed and higher solution
quality, of the hybrid algorithms (Li, Yang, & Nguyen, 2012; Huang, Liu, Su, & Yang, 2013). Meanwhile, detecting
strategies (Xia, Liu, & Hu, 2014) and opposition-based learning (OBL) strategies (Yu, Zhu, Ma, & Mao, 2015) are also
introduced to help the population to jump out of locally optimal solutions. Extensive experimental results manifest that
the comprehensive performance of a hybrid algorithm can be dramatically improved if various merits within different
algorithms/strategies are fully utilized by a proper integration mechanism (Xia, Gui, He, Xie, Wei, Xing, Wu, & Tang,
2018).

The explanatory information about the characteristics and principles of FA and PSO is mentioned in the following
subsections in order to make the novel hybrid method F-PSO more comprehensible.

Firefly Algorithm (FA)

FA is a relatively new nature-inspired optimization method that was first proposed by Yang (2010). As a stochastic
metaheuristic algorithm, FA incorporates randomness into a search process (Fister, Yang, Brest, & Fister, 2014). Even
though FA is similar to PSO, it has proved to be much simpler in algorithm implementation (Kora & Rama Krishna, 2016).
On the other hand, the method has proved effective in solving multimodal optimization problems (Yang & Wang, 2016).

According to FA process, some idealized conditions are adopted to search an optimal solution. These conditions are as
follows:
 The fireflies are attracted by other fireflies according to their brightness and distance between them.
 The attractiveness between two firefly colonies is proportionally increased with the brightness.
 The brightness of a firefly is associated with objectives and if there is no firefly brighter than the other, that firefly
will update.

In FA, it is assumed that there is a group of glowing fireflies. The attractiveness and the variation of the light intensity
are the most important factors to be considered to formulate the behavior of the fireflies. Here, the light intensity I is
307
proportional to the objective function of the problem desired to be optimized. The brightness of a firefly depends on
the intensity of light emitted by the firefly. The light intensity I varies with distance r which is expressed by the following
Eq. (7).

I(𝑟) = I0𝑒−𝛾𝑟2 (7)

where I0 denotes the intensity of the light at the source, and γ is a fixed light absorption coefficient .The distance 𝑟
depends on the attractiveness which is calculated according to the Eq. (8).

𝛽 (𝑟) = 𝛽0𝑒−𝛾𝑟2 (8)

where 𝛽0 is the largest attraction. All fireflies have a unique attractiveness 𝛽, which indicates the ability to attract other
fireflies. The attractiveness is related to the distance factor 𝑟𝑖𝑗 at locations s𝑖 and s𝑗, and between the two corresponding
fireflies, 𝑖 and 𝑗 are given by Eq. (9).

2
𝑟𝑖𝑗 = ǁ s𝑖- sj ǁ = √∑𝑑𝑘=1(𝑠𝑖𝑘 − 𝑠𝑗𝑘 ) (9)

where sik is the k-th element of the ith firefly position within the search space. Each firefly i move to another more
attractive firefly j and the movement is computed as in the Eq. (10).

1
sj =si + 𝛽0𝑒−𝛾𝑟𝑖𝑗2 (s𝑗 − s𝑖) + α(𝑟𝑎𝑛𝑑 − ) α ϵ [0, 1] (10)
2

Eq. (10) consists of three terms. The first term of the equation shows the position of the firefly i. The second term
demonstrates to the attractiveness, while the third term depends on the randomized move of the firefly i within the
search space. This term consists of the randomized parameter α, and randomly selected number between the interval
[0, 1] as rand adopted from a Gaussian distribution. The process of FA is demonstrated on a flowchart in Figure 1.

Figure 1. Flowchart of FA. Source: Self Elaboration.

START

Initialize the population

Calculate the fitness values

of fireflies

Update the intensity of

light

Update the position of fireflies

NO
Is termination criteria
met?

YES

STOP

308
Particle Swarm Optimization (PSO)

Particle swarm optimization (PSO) is a type of swarm based algorithm inspired by nature. On the basis of such
algorithms, living community called swarm has movements which are related to psychosocial factors. In PSO, each
individual is called a particle, and the interactions of particles create the swarm intelligence. The particles of a
community are better than the individual situation in terms of their observative and cognitive abilities. This status is
particularly seen in bees, birds, fish and even bacteria. In behaviors such as hunting, migrating and escaping, each
particle contributes to the consciousness of the swarm through its previous experiences and instincts. At this point,
some particles lead the swarm, while the other particles follow the pioneers as a harmonious member of the swarm
(Rajasekhar, Lynn, Das, & Suganthan, 2017).

PSO was introduced by Kennedy & Eberhart (1995) as a new optimization method in which each particle represents a
possible solution.

According to process of PSO, if the position and velocity of the particle i which is in the solution space at the moment of
t are shown xi(t) and vi(t) respectively, then the position of the particle i at the moment of (t+1) can be expressed as in
the Eq. (11).

xi(t+1) = xi(t) + vi(t+1) (11)

The component which is shown as the velocity vector in the expression is one of the main elements of the system. The
velocity vector includes information which is obtained by both its own and neighbor’s experiences. In PSO method, the
personal experiential knowledge is called cognitive component of the particle and the component obtained from
neighbors is also known as the social component. The position of each particle which represents a solution is updated
for every iteration in the algorithm. This update is made on the basis of each particle's best position lbest (best in the
locally) and the best of the swarm gbest (best in the globally) for each dimension j ϵ1, ..., N, where N is the dimension of
the problem. For this reason, it is very important to define the velocity vector correctly, which has both cognitive and
social components. Hence, vij represents the jth element of the velocity vector of the ith particle. Thus, the velocity of
particle i is updated by Eq. (12)

vij(t+1) = wvij(t) + c1r1j(t)[lbest(t) – xij(t)] + c2r2j(t)[gbest(t) – xij(t)] (12)

where w is the inertia weight which varies 0.8 to 1.2, c1 and c2∼U(0,2) are the acceleration coefficients, and also r1,
r2∼U(0,1) are stochastic random numbers. The inertia weight w influences the ability of the algorithm to search
solutions locally or globally. If the value of w decreases, then algorithm tends to search locally, but if w increases, globally
search is more possible. Selecting all the coefficients and parameters used in the PSO should be compatible with the
structure of the problem. This issue is quite essential in terms of facilitating the access of optimum.

The personal best position of particle i is the best point which is visited by particle i so far. Actually best position results
the best fitness value. Hereby the personal best of a particle at time t is updated as Eq. (13) where f denotes the objective
function that has to be minimized.

yi(t) , f(xi(t+1)) ≥ f(yi(t))

yi(t+1) = (13)
xi(t+1) , f(xi(t+1)) < f(yi(t))

In the Eq. 13, f function represents the fitness function. This function indicates that the position of a particle is close to
optimum or not. Eq. 14 is used to obtain the global best position (yg) from the individual best position.

yg(t)=min{f(xi(t)),...,f(xn(t))} (14)

A fundamental PSO algorithm is outlined in Figure 2.

309
 Figure 2. Flowchart of PSO. Source: Albayrak, 2017).

START

Initialize the positions and

velocities of each particle

Evaluate the fitness

value of each particle

Calculate lbest values of the

particles and gbest value of the
swarm

Update velocities of each

particle

Update positions of each

particle

Is termination criterion met?

NO
YES

STOP

Novel hybrid algorithm Firefly-Particle Swarm Optimization (F-PSO)

Combining two or more metaheuristic algorithms means enhancing the advantages of the algorithms to create the
hybrid algorithm that has practical value. Also the hybrid algorithm overcomes the shortcomings of the base algorithms
and widens the application areas.

In this research, the novel hybrid metaheuristic method named firefly-particle swarm optimization (F-PSO) is introduced
in order to solve TCTO problem. F-PSO is developed by modifying FA and PSO. Both algorithms have their own
advantages and wide range application areas. The movement of an individual is only depends on all other brighter
individuals in FA. Thus, the experimental knowledge of the individual can not affect the searching behavior. Therefore,
while the brightest individual is located in a local optimum, the population is easily trapped into premature convergence.
In contrast with FA, PSO takes into account historical knowledge of particles to guide them to search for promising
regions. Owing to the hybridization of these well-studied metaheuristic methods, F-PSO is characterized by high speed
and quick convergence properties. The specifications of F-PSO are described below in detail.

The algorithm of F-PSO which visualized in Figure 3 consists of two main phases. The algorithm implements FA in the
first phase of the method, and then PSO is applied in the second phase. FA works to refine the randomized population.
After first elimination, PSO handles the fireflies as particles to search the solution space and optimize the problem
globally.

In the hybrid algorithm, the elimination rate is accepted as 80%, after many experimental runs are performed (Albayrak,
2017). As a result of tests, this determined rate gives the most reasonable and consistent outputs. If this rate is increased
to over 90%, the elimination rate is not able to contribute to hybridization of the algorithms. Similarly, if the rate is
reduced fewer than 75%, the solution space affects negatively due to contraction. Consequently, the rate of elimination
is recommended in the range of 80% -85% for this type of problems. The phases of the algorithm are explained below.

Phase 1. FA is used to obtain the elite population for PSO.

(1.a). Initialize the firefly population which has n randomized individuals.
(1.b). Calculate the fitness values of each firefly by using objective function.
(1.c). Update the fireflies in according to light intensity and position.
310
 (1.d). Determine the top 0.8n in ranked population.
(1.e). Forward the elite population to PSO algorithm.

Phase 2. PSO is used to search the global best individual of the population.
(2.a). Initialize PSO algorithm considering elite population.
(2.b). Calculate the fitness values of particles.
(2.c). Find the global best position in the population.
(2.d). Update the particles in according to positions and velocities.
(2.e). Check whether the termination criteria are met; if so, terminate the algorithm; otherwise, follow the step
(1.b).

According to the algorithm flow, the fireflies explore larger search area initially. Thus, convergence speed of the
algorithm increases and PSO process performs global exploration more efficiently. The F-PSO algorithm is simply
structured and easy to use, while demonstrating great robustness and fast convergence in solving multi objective global
optimization problem.

Verification of F-PSO

In order to test and verify F-PSO, there are two common benchmark functions that can be used. These functions provide
a standardized comparison between PSO, FA, GA, ACO and F-PSO. Both reliability and accuracy abilities of the algorithms
can be investigated and compared through benchmark functions which are suitable with discrete and multimodal
optimization problems in the literature (Ma et al., 2014). Also the selected functions f1 and f2 are two of the most
competitive benchmarks due to many steep hollows with multiple local optimum points. The function specifications are
shown in Table 1.

Table 1. Specifications of benchmark functions. Source: Velásquez, 2010.

Function name Test function Variable domain Global optimum

Rastrigin (1974) f1(x) = ∑𝐷 2
𝑖=1(𝑥𝑖 − 10 𝑐𝑜𝑠(2𝜋𝑥𝑖 ) + 10) xi ϵ [-5.12, 5.12] xi = 0 , f1 (x) = 0
1 𝑥
Griewank (1981) f2(x) = ∑𝐷 2 𝐷 𝑖
𝑖=1 𝑥𝑖 − ∏𝑖=1 cos( ) + 1 xi ϵ [-5.12, 5.12] xi = 0 , f2 (x) = 0
4000 √𝑖

For the implementation of the benchmark functions, the experimental parameters of PSO, FA, GA, ACO and F-PSO are
given in Table 2.

Table 2. Experimental parameters of the algorithms. Source: Self elaboration.

Exp.
PSO FA GA ACO F-PSO
parameters
w 0.85 - - - 0.85
c 1.35 - - - 1.35
r 0.5 - - - 0.5
vmax 0.5 - - - 0.5
α - 0.2 - - 0.2
β - 0.2 - - 0.2
γ - 0.9 - - 0.9
δ - 0.95 - - 0.95
Number of
- - 10 - -
chromosome
single point and
Crossover - - double point
- -
Mutation - - single bit random - -
ρ - - - 0.8
ρo - - - 0.2 -
Search space
- - - [-20 20] -
limit
Pop. size 40 40 40 40 40
Max. iteration 1000 1000 1000 1000 1000

311
Searching for the minimum value of the functions, 40 experiments are carried out for each method which is limited
1000 iterations. The results are summarized in Table 3 included best, mean and standard deviation values. As can be
seen from Tab. 3, the results show that F-PSO outperforms the results of standard algorithms. In other words, F-PSO is
the best method in terms of solving the functions f1 and f2 which are highly multimodal.

The exploitation and exploration are two major factors among the many components of PSO and FA algorithms in order
to search the solution space on a global scale efficiently. The enhanced searching ability of F-PSO is based on
consolidation of powerful algorithms properly. The two-stage structure of the hybrid algorithm prevents to get stuck in
local optimum which is the main shortcoming of the base algorithms. The prior knowledge obtained FA guides global
search process for PSO stage. Hereby, better solution candidates are left to the secondary process.

Figure 3. Flowchart of F-PSO. Source: Self Elaboration.

Table 3. Test results of the methods. Source: Self Elaboration.

Function name Method Best value Mean value Std. deviation
Rastrigin PSO 0.8351 0.9835 0.1211
FA 0.6433 0.8821 0.1953
GA 0.7706 0.9145 0.1738
ACO 0.6179 0.7224 0.1244
F-PSO 0.5211 0.6374 0.0936
Griewank PSO 0.0228 0.0454 0.0185
FA 0.0723 0.0889 0.0249
GA 0.0812 0.1003 0.0209
ACO 0.0798 0.0926 0.0227
F-PSO 0.0197 0.0256 0.0168

Application of Time-Cost Trade-Off (TCTO)

The application of F-PSO is examined on a sample project which includes TCTO problem. The sample project is selected
from the literature that enables comparison. This 18-activity network and time-cost components were described in
(Feng et al., 1997). Multiplying the number of execution modes with 18 activities by each other, there are approximately
5.9 billion combinations in the entire project.
312
For this implementation, F-PSO procedures are coded MATLAB R2012a. Then, the algorithm is run on a personal
computer configured with Intel Core2, 4 GB RAM, Windows10.

In order to observe the coherence of the solutions, the execution is repeated 20 times (Albayrak, 2017). Because the
true Pareto front (i.e. non-dominated solutions) is difficult to be generated by a few run.

Application Example

The project is visualized with an activity-on-node (AON) network in Figure 4. Hereby, the precedence relationships can
be seen in this figure easily. In sample project, each activity has different execution modes corresponding time and cost
options. The activity numbers, predecessors, durations and cost alternatives are presented in Table 4. Along with five
execution modes of construction, duration varies from faster to slower and similarly, cost varies from cheaper to more
expensive. Additionally, the problem is considered to be minimized according to total project duration and cost.

Figure 4. AON network of the problem. Source: Adapted from Feng et al. (1997).

1 5 7 11
8
6 1
9 1
7
Star 2 10 1 1
t 5

3 13

4 14 1

Table 4. 18-activity TCTO problem. Source: Feng et al. (1997).

Act. Pred.
Mode 1 Mode 2 Mode 3 Mode 4 Mode 5
No. Activity
Duration Cost Duration Cost Duration Cost Duration Cost Duration Cost
(days) ($) (days) ($) (days) ($) (days) ($) (days) ($)
1 - 14 2400 15 2150 16 1900 21 1500 24 1200
2 - 15 3000 18 2400 20 1800 23 1500 25 1000
3 - 15 4500 22 4000 33 3200 - - - -
4 - 12 45000 16 35000 20 30000 - - - -
5 1 22 20000 24 17500 28 15000 30 10000 - -
6 1 14 40000 18 32000 24 18000 - - - -
7 5 9 30000 15 24000 18 22000 - - - -
8 6 14 220 15 215 16 210 21 208 24 120
9 6 15 300 18 240 20 180 23 150 25 100
10 2, 6 15 450 22 400 33 320 - - - -
11 7, 8 12 450 16 350 20 300 - - - -
12 5, 9, 10 22 2000 24 1750 28 1500 30 1000 - -
13 3 14 4000 18 3200 24 1800 - - - -
14 4, 10 9 3000 15 2400 18 2200 - - - -
15 12 12 4500 16 3500 - - - - - -
16 13, 14 20 3000 22 2000 24 1750 28 1500 30 1000
17 11, 14, 15 14 4000 18 3200 24 1800 - - - -
18 16, 17 9 3000 15 2400 18 2200 - - - -

Albayrak (2017) performed many preliminary experiments and suggested the number of iterations as 500. Similarly, the
population size was proposed as 20 for this problem. These suggestions are adopted for this paper. To perform the
problem with F-PSO, the parameters of the algorithm are set as Table 5.

313
 Table 5. F-PSO experiments for application. Source: Self Elaboration.
Parameters of F-PSO
w 0.85
c 1.35
vmax 0.5
α 0.2
β 0.2
γ 0.9
δ 0.95
Pop. size 20
Max. iteration 500

Results

The FA algorithm starts with random generation of the initial population. Each obtained value is represented by a firefly
which is a candidate of the optimum result. The light intense of fireflies indicates the fitness value according to the
objective function. Then fireflies are compared with each other and the best 80% of the population is transferred to
PSO process. The position and velocity update operators are applied to population repeatedly until the termination
condition is satisfied. The best global optimal value is provided by the end of the procedure.

The method is able to search the solution space and archive the dominated and non-dominated solutions in the
successive iterations of the algorithm. After the analysis, the dominated solutions are eliminated and non-dominated
solutions are kept. This leads to separate 24 Pareto optimal solutions for this project. Each of these solutions correspond
an optimal trade-off point between total time and cost. Pareto front solutions are summarized in Table 6 and Figure 5
shows the curve of the time-cost trade-off relating to this project, where the horizontal axis represents total duration
and the vertical axis represents project cost. In according to TCTO curve, the project duration varies from 100 to 152
days and the cost values vary from $134.870,00 to $100.240,00 totally. Each of Pareto solutions represents an optimal
trade-off for the activities of the project entirely.

Table 6. Pareto solutions of TCTO. Source: Self Elaboration.

Solution No. Duration (days) Cost ($x103) Solution No. Duration (days) Cost ($x103)
1 100 134.87 13 122 104.02
2 103 129.8 14 123 103.85
3 104 126.9 15 124 102.7
4 105 121.12 16 125 102.1
5 108 116.415 17 126 101.87
6 111 114.09 18 128 101.05
7 113 111.915 19 131 100.98
8 115 109.9 20 135 100.75
9 118 107.52 21 139 100.66
10 119 106.27 22 143 100.52
11 120 104.92 23 147 100.378
12 121 104.77 24 152 100.24

Figure 5. Pareto front of TCTO. Source: Self Elaboration.

140
130
Cost ($ x 103)

120
110
100
90
90 100 110 120 130 140 150 160
Duration (days)

314
The Pareto front generated by the proposed model provides critical information needed by decision makers to create
optimal designs under project time and cost constraints. The proposed model allows decision makers to make timely,
informed decisions based on time and cost preferences.

In F-PSO, which is limited to 500 iterations, the values forming the Pareto front were observed in the first 200 iterations.
The graphic of number of iterations-fitness value is given in the Figure 6. According to the Figure 6, it is concluded that
the selected parameters and the number of iterations are sufficient.

Figure 6. Convergence of algorithm in terms of fitness value and iterations. (Source: Self Elaboration)Source:

140
Fitness value (x103)

130
120
110
100
90
0 100 200 300 400 500
No. of iterations

In order to validate the results, Elbeltagi et al. (2005), Geem (2010) and Lee et al. (2015) are considered in the literature
because 18-activity project is examined by them using other metaheuristic algorithms. The algorithms to be compared
are as follows: Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Genetic Algorithm (GA), Memetic
Algorithm (MA), Harmony Search (HS) and Advanced Stochastic Time-Cost Tradeoff (ASTCT). The comparison confirms
that the current model is capable of obtaining better solution in terms of both cost and duration results. According to
comparison, harmony search optimal values are second optimal among seven metaheuristic algorithms. The results are
compared visually as shown in Table 7.

Table 7. Comparison of seven metaheuristic algorithm results on the same problem. Source: Self Elaboration.

(Elbeltagi et al., 2005) (Geem, 2010) (Lee et al., 2015) This study
PSO ACO GA MA HS ASTCT F-PSO
Min. Cost ($) 106270 106270 105770 106270 102320 106270 100240
Min. Dur.(days) 110 110 113 110 105 110 100

Conclusions

Every construction project has its own unique conditions. The uniqueness of construction projects also means that the
external influences and constraints would be different, yet subject to change throughout the project timeline.
Therefore, it is important to attain time-cost alternatives that are appropriate to the project conditions. This approach
can be achieved through metaheuristics and Pareto front.

This paper presents a novel method named F-PSO for solving TCTO problem which is one of the most difficult discrete
optimization problems in construction project planning, due to various variables and constraints. F-PSO is based on the
hybridization of FA and PSO to improve the exploration and exploitation abilities of both algorithms. Also, better
optimality of the solutions and quick convergence are aimed when building the hybrid method. For this purpose, the
poor solutions according to fitness values are disqualified in FA section of the process, before passing PSO that is the
second step of the algorithm. Primarily, two benchmark equations are tested to verify the algorithm. After verification,
F-PSO is implemented on a well-known 18-activity TCTO problem which enables the comparison. The proposed
algorithm finds out the optimal solution and defines the Pareto front. The results are compared to other five
metaheuristic algorithms which are applied on the same problem. The comparison shows that, F-PSO is generally found
to perform better than other algorithms in terms of both minimum cost and duration results. The improvement of the
algorithm combination provides an efficient method regarding to obtain shorter and more economical alternatives of
the construction projects.

315
The proposed F-PSO is simple, robust and efficient. It does not impose any limitation on the number of objectives and
can be extended to include more objectives. Further minor modifications of F-PSO algorithm hold potential to resolve
other multi-objective optimization problems in the field of construction management such as the tradeoffs among
performance, cost, and reliability; time, cost and safety tradeoffs; and resource-constrained and resource-leveling in
project scheduling activities.

In this study, the quality and risk factors are not considered and they are assumed stable against the changes of time
and cost. The idea of integrating such factors into study will provide new insights for future works.

Acknowledgement

This paper is based on the first author’s PhD dissertation entitled “A Comparative Selection of Hybrid-Metaheuristic
Methods on Time-Cost Trade-Off in Resource Constrained Construction Projects”.

The authors would like to express their sincere thanks to the anonymous reviewers for valuable comments and
suggestions on this paper.

References

Alaghebandha, M., Hajipour, V., & Hemmati, M. (2017). Optimizing multi-objective sequencing problem in mixed-model assembly line on just-in-time:
particle swarm optimization algorithm. International Journal of Management Science and Engineering Management, 12(4), 288–298.
https://doi.org/10.1080/17509653.2016.1258593

Albayrak, G. (2017). A comparative selection of hybrid-metaheuristic methods on time-cost trade-off in resource constrained construction projects.
PhD Dissertation, Eskisehir Osmangazi University, Graduate School of Natural and Applied Sciences, Turkey.

Ashuri, B., & Tavakolan, M. (2015). Shuffled Frog-Leaping Model for Solving Time-Cost-Resource Optimization Problems in Construction Project
Planning. Journal of Computing in Civil Engineering, 29(1), 04014026. https://doi.org/10.1061/(ASCE)CP.1943-5487.0000315

Blum, C., Belsa Aguilera, M. J., Roli, A., & Sampels, M. (2013). Hybrid Metaheuristics. Computational Intelligence (Vol. 7919).
https://doi.org/10.1007/978-3-642-38516-2

Chen, S. P., & Tsai, M. J. (2011). Time-cost trade-off analysis of project networks in fuzzy environments. European Journal of Operational Research,
212(2), 386–397. https://doi.org/10.1016/j.ejor.2011.02.002

Cheng, M. Y., & Tran, D. H. (2014). Two-phase differential evolution for the multiobjective optimization of time-cost tradeoffs in resource-constrained
construction projects. IEEE Transactions on Engineering Management, 61(3), 450–461. https://doi.org/10.1109/TEM.2014.2327512

De, P., Dunne, E. J., Ghosh, J. B., & Wells, C. E. (1997). Complexity of the Discrete Time-Cost Tradeoff Problem for Project Networks. Operations
Research, 45(2), 302–306. https://doi.org/10.1287/opre.45.2.302

Deǐneko, V. G., & Woeginger, G. J. (2001). Hardness of approximation of the discrete time-cost tradeoff problem. Operations Research Letters, 29(5),
207–210. https://doi.org/10.1016/S0167-6377(01)00102-X

Elbeltagi, E., Hegazy, T., & Grierson, D. (2005). Comparison among five evolutionary-based optimization algorithms. Advanced Engineering Informatics,
19(1), 43–53. https://doi.org/10.1016/j.aei.2005.01.004

Elloumi, S., Fortemps, P., & Loukil, T. (2017). Multi-objective algorithms to multi-mode resource-constrained projects under mode change disruption.
Computers and Industrial Engineering, 106, 161–173. https://doi.org/10.1016/j.cie.2017.01.029

Elmaghraby, S. E., & Salem, A. (1981). Optimal linear approximation in project compression. Operations Research Technical Report, 171, North Carolina
State University at Raleigh.

Eshtehardian, E., Afshar, A., & Abbasnia, R. (2009). Fuzzy-based MOGA approach to stochastic time-cost trade-off problem. Automation in
Construction, 18(5), 692–701. https://doi.org/10.1016/j.autcon.2009.02.001

Fan, K., You, W., & Li, Y. (2013). An effective modified binary particle swarm optimization (mBPSO) algorithm for multi-objective resource allocation
problem (MORAP). Applied Mathematics and Computation, 221, 257–267. https://doi.org/10.1016/j.amc.2013.06.039

Feng, C.-W., Liu, L., & Burns, S. A. (1997). Using Genetic Algorithms to Solve Construction Time-Cost Trade-Off Problems. Journal of Computing in Civil
Engineering, 11(3), 184–189. https://doi.org/10.1061/(ASCE)0887-3801(1997)11:3(184)

Feng, C.-W., Liu, L., & Burns, S. A. (2000). Stochastic Construction Time-Cost Trade-Off Analysis. Journal of Computing in Civil Engineering, 14(2), 117–
126. https://doi.org/10.1061/(ASCE)0887-3801(2000)14:2(117)

Fister, I., Yang, X. S., Brest, J., & Fister, I. Jr. (2014). On the randomized firefly algorithm. Cuckoo Search and Firefly Algorithm. Springer International
Publishing, 27–48.

Fulkerson, D. R. (1961). A Network Flow Computation for Project Cost Curves. Management Science, 7(2), 167–178.
https://doi.org/10.1287/mnsc.7.2.167

316
Geem, Z. W. (2010). Multiobjective Optimization of Time-Cost Trade-Off Using Harmony Search. Journal of Construction Engineering and
Management, 136(6), 711–716. https://doi.org/10.1061/(ASCE)CO.1943-7862.0000167

Griewank, A. O. (1981). Generalized descent for global optimization. Journal of Optimization Theory and Applications, 34(1), 11–39.
https://doi.org/10.1007/BF00933356

He, Z., He, H., Liu, R., & Wang, N. (2017). Variable neighbourhood search and tabu search for a discrete time/cost trade-off problem to minimize the
maximal cash flow gap. Computers and Operations Research, 78, 564–577. https://doi.org/10.1016/j.cor.2016.07.013

Hegazy, T. (1999). Optimization of construction time-cost trade-off analysis using genetic algorithms. Canadian Journal of Civil Engineering, 26(6),
685–697. https://doi.org/10.1139/l99-031

Hendrickson, C., & Au, T. (1989). Project management for construction. Prentice Hall, Englewood Cliffs, N.J., USA.

Hou, L., Zhao, C., Wu, C., Moon, S., & Wang, X. (2017). Discrete Firefly Algorithm for Scaffolding Construction Scheduling. Journal of Computing in Civil
Engineering, 31(3), 04016064. https://doi.org/10.1061/(ASCE)CP.1943-5487.0000639

Huang, S. J., Liu, X. Z., Su, W. F., & Yang, S. H. (2013). Application of hybrid firefly algorithm for sheath loss reduction of underground transmission
systems. IEEE Transactions on Power Delivery, 28(4), 2085–2092. https://doi.org/10.1109/TPWRD.2013.2265913

Huang, Y. S., Deng, J. J., & Zhang, Y. Y. (2008). TI time-cost-quality tradeoff optimization in construction project based on modified ant colony algorithm.
In Proceedings of the 7th International Conference on Machine Learning and Cybernetics, ICMLC (Vol. 2, pp. 1031–1035).
https://doi.org/10.1109/ICMLC.2008.4620556

Kandil, A., & El-Rayes, K. (2005). Parallel Computing Framework for Optimizing Construction Planning in Large-Scale Projects. Journal of Computing in
Civil Engineering, 19(3), 304–312. https://doi.org/10.1061/(ASCE)0887-3801(2005)19:3(304)

Kaveh, A., Khanzadi, M., Alipour, M., & Naraky, M. R. (2015). CBO and CSS algorithms for resource allocation and time-cost trade-off. Periodica
Polytechnica Civil Engineering, 59(3), 361–371. https://doi.org/10.3311/PPci.7788

Kelley, J. E. (1961). Critical-Path Planning and Scheduling: Mathematical Basis. Operations Research, 9(3), 296–320.
https://doi.org/10.1287/opre.9.3.296

Kennedy, J., & Eberhart, R. (1995). Particle swarm optimization. Proceedings of IEEE International Conference on Neural Networks, 1942–1948.
https://doi.org/10.1109/ICNN.1995.488968

Klerides, E., & Hadjiconstantinou, E. (2015). The stochastic discrete time-cost tradeoff problem with decision-dependent uncertainty. In Handbook on
Project Management and Scheduling Vol. 2 (pp. 781–809). https://doi.org/10.1007/978-3-319-05915-0_6

Kora, P., & Rama Krishna, K. S. (2016). Hybrid firefly and Particle Swarm Optimization algorithm for the detection of Bundle Branch Block. International
Journal of the Cardiovascular Academy, 2(1), 44–48. https://doi.org/10.1016/j.ijcac.2015.12.001

Lee, H. G., Yi, C. Y., Lee, D. E., & Arditi, D. (2015). An Advanced Stochastic Time-Cost Tradeoff Analysis Based on a CPM-Guided Genetic Algorithm.
Computer-Aided Civil and Infrastructure Engineering, 30(10), 824–842. https://doi.org/10.1111/mice.12148

Leu, B. S., & Yang, C. (1999). GA-Based Multicriteria Optimal for Construction Scheduling. Journal of Construction Engineering and Management,
125(December), 420–427. https://doi.org/10.1061/(ASCE)0733-9364(1999)125:6(420)

Li, C., Yang, S., & Nguyen, T. T. (2012). A self-learning particle swarm optimizer for global optimization problems. IEEE Transactions on Systems, Man,
and Cybernetics, Part B: Cybernetics, 42(3), 627–646. https://doi.org/10.1109/TSMCB.2011.2171946

Li, H., & Love, P. (1997). Using Improved Genetic Algorithms to Facilitate Time-Cost Optimization. Journal of Construction Engineering and
Management, 123(3), 233–237. https://doi.org/10.1061/(ASCE)0733-9364(1997)123:3(233)

Ma, L., Hu, K., Zhu, Y., Niu, B., Chen, H., & He, M. (2014). Discrete and continuous optimization based on hierarchical artificial bee colony optimizer.
Journal of Applied Mathematics, 2014 Article ID 402616, 20 pages. https://doi.org/10.1155/2014/402616

Mehdizadeh, E., Hajipour, V., & Mohammadizadeh, M. R. (2015). A bi-objective multi-item capacitated lot-sizing model: Two pareto-based meta-
heuristic algorithms. International Journal of Management Science and Engineering Management, 11(4).
https://doi.org/10.1080/17509653.2015.1086965

Mokhtari, H., Baradaran Kazemzadeh, R., & Salmasnia, A. (2011). Time-cost tradeoff analysis in project management: An ant system approach. IEEE
Transactions on Engineering Management, 58(1), 36–43. https://doi.org/10.1109/TEM.2010.2058859

Ng, S. T., & Zhang, Y. (2008). Optimizing Construction Time and Cost Using Ant Colony Optimization Approach. Journal of Construction Engineering
and Management, 134(September), 721–728. https://doi.org/10.1061/(ASCE)0733-9364(2008)134:9(721)

Pathak, B. K., & Srivastava, S. (2015). Effects of project uncertainties on nonlinear time-cost tradeoff profile. Iranian Journal of Fuzzy Systems, 12(4),
79–100.

Pareto, V. (1906). Manuale di economia politica, societa editrice libraria. Milano, Italy. Translated into English by Schwier, A.S. (1971) Manual of
political economy, New York: Macmillan.

Praščević, N. & Praščević, Ž. (2014). Application of particle swarms for project time-cost optimization, GRAĐEVINAR, 66(12), 1097-1107, doi:
https://doi.org/10.14256/JCE.1048.2014

Rajasekhar, A., Lynn, N., Das, S., & Suganthan, P. N. (2017). Computing with the collective intelligence of honey bees – A survey. Swarm and
Evolutionary Computation, 32, 25-48. https://doi.org/10.1016/j.swevo.2016.06.001

317
Rastrigin, L.A. (1974). Extremal Control Systems. In Theoretical Foundations of Engineering Cybernetics Series, Moscow.

Salem, A. M., & Elmaghraby, S. E. (1994). Optimal linear approximation in project compression. IIE Transactions (Institute of Industrial Engineers),
16(4), 339–347. https://doi.org/10.1080/07408178408975253

Shelokar, P. S., Siarry, P., Jayaraman, V. K., & Kulkarni, B. D. (2007). Particle swarm and ant colony algorithms hybridized for improved continuous
optimization. Applied Mathematics and Computation, 188(1), 129–142. https://doi.org/10.1016/j.amc.2006.09.098

Shen, Z., Hassani, A., & Shi, Q. (2016). Multi-objective time-cost optimization using Cobb-Douglas production function and hybrid genetic algorithm.
Journal of Civil Engineering and Management, 22(2), 187–198. https://doi.org/10.3846/13923730.2014.897966

Siemens, N. (1971). A Simple CPM Time-Cost Tradeoff Algorithm. Management Science, 17(6), B-354-B-363. https://doi.org/10.1287/mnsc.17.6.B354

Sonmez, R., & Bettemir, Ö. H. (2012). A hybrid genetic algorithm for the discrete time-cost trade-off problem. Expert Systems with Applications, 39(13),
11428–11434. https://doi.org/10.1016/j.eswa.2012.04.019

Sörensen K., Sevaux M., & Glover F. (2018). A history of metaheuristics. In: Martí R., Panos P., Resende M. (eds) Handbook of Heuristics. Springer,
Cham. https://doi.org/10.1007/978-3-319-07153-4_4-1

Tareghian, H. R., & Taheri, H. (2006). An application of randomized minimum cut to the project time/cost tradeoff problem. Applied Mathematics and
Computation, 173(2), 1200–1207. https://doi.org/10.1016/j.amc.2005.04.063

Tareghian, H. R., & Taheri, S. H. (2007). A solution procedure for the discrete time, cost and quality tradeoff problem using electromagnetic scatter
search. Applied Mathematics and Computation, 190(2), 1136–1145. https://doi.org/10.1016/j.amc.2007.01.100

Vanhoucke, M., & Debels, D. (2007). The discrete time/cost trade-off problem: Extensions and heuristic procedures. Journal of Scheduling, 10(4–5),
311–326. https://doi.org/10.1007/s10951-007-0031-y

Velásquez, J. D. (2010). An Enhanced Hybrid Chaotic Algorithm using Cyclic Coordinate Search and Gradient Techniques. Revista de Ingeniería,
Universidad de Los Andes, 32, 45–53.

Xia, X., Gui, L., He, G., Xie, C., Wei, B., Xing, Y., … Tang, Y. (2018). A hybrid optimizer based on firefly algorithm and particle swarm optimization
algorithm. Journal of Computational Science, 26, 488-500. https://doi.org/10.1016/j.jocs.2017.07.009

Xia, X., Liu, J., & Hu, Z. (2014). An improved particle swarm optimizer based on tabu detecting and local learning strategy in a shrunk search space.
Applied Soft Computing Journal, 23, 76–90. https://doi.org/10.1016/j.asoc.2014.06.012

Yang, X. S. (2010). Firefly algorithm, stochastic test functions and design optimisation, International Journal of Bio-Inspired Computation, 2(2), 78-84.

Yang, X. S., Deb, S., & Fong, S. (2014). Metaheuristic algorithms: Optimal balance of intensification and diversification. Applied Mathematics and
Information Sciences, 8(3), 977–983. https://doi.org/10.12785/amis/080306

Yang, Z., & Wang, J. (2016). Multistep wind speed forecasting using a novel model hybridizing singular spectrum analysis, modified intelligent
optimization, and rolling elman neural network. Mathematical Problems in Engineering, 2016, Article ID 3623412, 21 pages.
https://doi.org/10.1155/2016/3623412

Yu, S., Zhu, S., Ma, Y., & Mao, D. (2015). Enhancing firefly algorithm using generalized opposition-based learning. Computing, 97(7), 741–754.
https://doi.org/10.1007/s00607-015-0456-7

Zheng, H., & Zhong, L. (2017). Discrete time-cost-environment trade-off problem and its application to a large-scale construction project. Proceedings
of the Tenth International Conference on Management Science and Engineering Management (pp.1375-1382). https://doi.org/10.1007/978-
981-10-1837-4_111

318