Int. J. Mathematics in Operational Research, Vol. 9, No.

1, 2016 125

Hybrid bat algorithm for flow shop scheduling

problems

Ömür Tosun*
Department of International Trade and Logistics,
Ayşe Sak School of Applied Sciences,
Akdeniz University, Yeşilbayır, Antalya, Turkey
Fax: +90-242-227-44-54
Email: [email protected]
*Corresponding author

M.K. Marichelvam
Department of Mechanical Engineering,
Mepco Schlenk Engineering College,
Sivakasi, Tamilnadu, 626005, India
Email: [email protected]

Abstract: In this paper, a hybrid bat algorithm is proposed to solve the

permutation flow shop scheduling problems with makespan criterion.
Developed for numerical optimisation, local search techniques are integrated
into the bat algorithm to solve the flow shop problems. Reeves test
problems from the literature are used to compare the performance of the
proposed algorithm against the best known solutions, NEH and with different
metaheuristics. Computational results and statistical analysis indicate that the
proposed algorithm can be more efficient than many other metaheuristic
algorithms addressed in the literature.

Keywords: flow shop; scheduling; makespan; bat algorithm; metaheuristics.

Reference to this paper should be made as follows: Tosun, Ö. and

Marichelvam, M.K. (2016) ‘Hybrid bat algorithm for flow shop scheduling
problems’, Int. J. Mathematics in Operational Research, Vol. 9, No. 1,
pp.125–138.

Biographical notes: Ömür Tosun is an Assistant Professor in the Department

of International Trade and Logistics in Ayse Sak School of Applied Sciences.
He holds a PhD in Production Management from Department of Business
Administration in the Faculty of Economic and Administrative Sciences,
Akdeniz University, Turkey. He also holds a BS in Industrial Engineering and
MBA in Production Management. He has an interest on neural networks,
optimisation, metaheuristic methods and bio-inspired algorithms.

M.K. Marichelvam is currently working as an Assistant Professor in the

Department of Mechanical Engineering at Mepco Schlenk Engineering
College, Sivakasi, Tamilnadu, India. He received his BE in Mechanical
Engineering from Madurai Kamaraj University in 2000 and his ME in
Industrial Engineering from Thiagarajar College of Engineering, Madurai,
India in 2002. His area of interest is manufacturing scheduling, multi-objective

Copyright © 2016 Inderscience Enterprises Ltd.

126 Ö. Tosun and M.K. Marichelvam

optimisation, heuristics and hybrid meta-heuristics. Currently, he is pursuing

his PhD at Anna University, Chennai, Tamilnadu, India. He has published more
than ten papers in referred journals and conferences.

1 Introduction

Flow shop scheduling is one of the most prevalent problems in the field of deterministic
scheduling in engineering industries. Permutation flow shop scheduling (PFSS) problem
is a specific case of flow shop scheduling problems which is widely used both in
academic and engineering applications. In this case, solutions are represented by the
permutations of n jobs, i.e., π = {π1, π2,…,πn}. Each job comprises a set of m operations
that must be performed by different machines. Each machine can process only one
operation at a time. Although there are exact solutions for one or two machines, the
problem becomes NP-hard for three or more machines. This challenging environment
attracts many researchers to determine more efficient solution algorithms based on
heuristics or metaheuristics.
Different performance criteria were considered by the researchers on PFSS such as
makespan, total flow time, earliness, tardiness or lateness of jobs. For solving the PFSS
problems many researchers developed their own solution approaches. These methods can
be broadly classified into three categories such as exact methods, constructive heuristic
methods [such as Johnson (1954), CDS (Campbell et al., 1970) and NEH (Nawaz et al.,
1983)] and metaheuristic algorithms. Most of the recent studies about the PFSS include
metaheuristics, that is, improvement algorithms that start from a certain initial sequence
and apply general-purpose local search procedures.
Bat algorithm (BA) is a relatively new nature-inspired and population-based
algorithm proposed by Yang (2010) inspired by the echolocation of microbats. Specially
developed for solving constrained optimisation problems, it is now used widely in
different types of combinatorial problems.
In this paper, a hybrid bat algorithm is used to solve the PFSS problems with
makespan (Cmax) criterion. Different benchmark problems addressed in the literature are
selected to evaluate the performance of the proposed algorithm. To the best of our
knowledge, this study is one of the first applications of bat algorithm for solving the
PFSS with makespan performance measure and also the first with a variety of benchmark
problems. The performance of the proposed algorithm is compared with different
metaheuristics addressed in the literature.
The rest of the paper is organised as follows. The FSSP is presented in Section 2.
Section 3 includes the bat algorithm and is followed by the proposed bat algorithm for
PFSS with makespan criterion in Section 4. The computational results and statistical
analysis are presented in Section 5. Concluding remarks and future scope of the algorithm
are given in Section 6.
 Hybrid bat algorithm for flow shop scheduling problems 127

2 PFSS problem

The PFSS problem, which is characterised as Fm/prmu/Cmax attracted the researchers for
the past several decades. In, the three field notation, the first term refers to the machine
environment (flow shop), second term is the processing restrictions or constraints
(permutation) and third term is the objective to be minimised (makespan) (Tosun, 2014).
In the flow shop problem, n jobs are to be processed on m machines. All jobs must be
processed on every machine in the same sequence. In this study, the objective is to find
the permutation of jobs that minimises the makespan. Makespan is defined as the
completion time of the last job in the production system. Makespan is a performance
measure used to improve the efficiency of the production system.
Let C(ji, k) denote the completion time of job ji on machine k, π = (j1, j2,…,jn) denote
a permutation of all jobs and pi,j is the process time of job i on machine j. The completion
time for an n-job, m-machine flow shop problem is calculated as follows (Zobolas et al.,
2009; Marichelvam, 2012):
C ( j1 , 1) = p j1 ,1 (1)

C ( ji , 1) = C ( ji −1 , 1) + p j1 ,1 , i = 2,… , n (2)

C ( j1 , k ) = C ( ji , k − 1) + p j1 , k , k = 2,… , m (3)

C ( ji , k ) = max {C ( ji −1 , k ) , C ( ji , k − 1)} + p ji , k , i = 2,… , n; k = 2,… , m (4)

Equation (1) gives the completion time of the first job of the permutation on machine 1.
Equation (2) is the completion time of the permutation on machine 1, equation (3) is the
completion time of the first job of the permutation on machine k and equation (4) is the
completion time of the permutation on machine k.
Makespan is given by:
Cmax (π ) = C ( jn , m ) (5)

So, the PFSS problem with the makespan criterion is to find a permutation π* in the set of
all permutations Π such that

Cmax ( π * ) ≤ C ( jn , m ) ∀π ∈ Π (6)

Researchers proposed many heuristics for solving the PFSS problems. However the
heuristics are problem specific. Hence, researchers developed many new solution
methods (called metaheuristics) and their results are compared with each other and also
with the more traditional methods (Framinan et al., 2005, 2004; Ying, 2012; Wang et al.,
2013).
Many researchers proposed different algorithms for different scheduling problems
with different objective functions (Mittal and Kanda, 2009; Shang and McKendall, 2009;
Najid and Arroub, 2010; Sharma et al., 2013).
128 Ö. Tosun and M.K. Marichelvam

Nature-inspired algorithms are based on the modelling the collective behaviour of

self-organised systems and applying these models to solve the real-world problems which
also known as swarm intelligence. In this study, a new solution approach based on bat
algorithm is proposed.

3 Bat algorithm

The standard bat algorithm was based on the echolocation or bio-sonar characteristics of
microbats. Microbats use a type of sonar called echolocation to detect prey, avoid
obstacles, and locate their roosting crevices in the dark. These bats emit a very loud
sound pulse and listen for the echo that bounces back from the surrounding objects. Their
pulses vary in properties and can be correlated with their hunting strategies, depending on
the species. Most bats use short, frequency-modulated signals to sweep through about an
octave, while others more often use constant-frequency signals for echolocation. Their
signal bandwidth varies depends on the species, and often increased by using more
(Yang, 2013; Marichelvam et al., 2013; Gandomi and Yang, 2014; Sabba and Chikhi,
2014). Yang (2011) also proposed the multi-objective BA to solve the multi-objective
design problems such as welded beam design.

3.1 Echolocation of microbats

Each pulse only lasts a few thousandths of a second (up to about 8 to 10 ms). However, it
has a constant frequency which is usually in the region of 25 kHz to 150 kHz. The typical
range of frequencies for most bat species are in the region between 25 kHz and 100 kHz,
though some species can emit higher frequencies up to 150 kHz. Each ultrasonic burst
may last typically 5 to 20 ms, and microbats emit about 10 to 20 such sound bursts every
second. When hunting for prey, the rate of pulse emission can be speed up to about
200 pulses per second when they fly near their prey. Such short sound bursts imply the
fantastic ability of the signal processing power of bats. In fact, studies shows the
integration time of the bat ear is typically about 300 to 400 μs. As the speed of sound in
air is 340 m/s, the wave length of the ultrasonic sound bursts with a constant frequency.
The wavelength is calculated as follows.
v
λ= (7)
f

where v is the speed of sound in air and f is the frequency. The wave length is in the
range of 2 mm to 14 mm for the typical frequency range from 25 kHz to 150 kHz. Such
wavelengths are in the same order of their prey sizes. Studies show that microbats use the
time delay from the emission and detection of the echo, the time difference between their
two ears, and the loudness variations of the echoes to build up three dimensional scenario
of the surrounding. However, here we are only interested in the echolocation and the
associated behaviour. Such echolocation behaviour of microbats can be formulated in
such a way that it can be associated with the objective function to be optimised, and this
makes it possible to formulate new optimisation algorithms.
 Hybrid bat algorithm for flow shop scheduling problems 129

3.2 Origins of the bat algorithm

If the echolocation characteristics of microbats can be idealised, one can develop various
bat-inspired algorithms or BAs. The following approximate or idealised rules were used
by Yang (2010).
1 All bats use echolocation to sense distance, and they also ‘know’ the difference
between food/prey and background barriers in some magical way.
2 Bats fly randomly with velocity vi at position xi with a frequency fmin, varying wave
length λ and loudness A0 to search for prey. They can automatically adjust the wave
length (or frequency) of their emitted pulses and adjust the rate of pulse emission
r ∈ [0, 1], depending on the proximity of their target.
3 Although the loudness can vary in many ways, we assume that the loudness varies
from a large (positive) A0 to a minimum constant value Amin.
For simplicity no ray tracing is used in estimating the time delay and three dimensional
topography. In addition to these assumptions, the following approximations are proposed.
In general the frequency f in a range (fmin, fmax) corresponds to a range of wave lengths
(λmin, λmax). For example a frequency range of (20 kHz, 500 kHz) corresponds to a range
of wave lengths from 0.7 mm to 17 mm in reality. Obviously, one can choose the ranges
freely to suit different applications.

3.3 Bat motion

Each bat is associated with a velocity vit and a location xit , at iteration t, in a
d-dimensional search or solution space. Among all the bats, there exists a current best
solution x*. Therefore, the above three rules can be translated into the updating equations
for xit and vit :

fi = f min + ( f max − f min ) β (8)

vit = vit −1 + ( xit −1 − x* ) fi (9)

xit = xit −1 + vit (10)

where β ∈ [0, 1] is a random vector drawn from a uniform distribution.

As mentioned earlier, one can either use wave lengths or frequencies for
implementation, in this study fmin = 0 and fmax = O(1) will be used, depending on the
domain size of the problem of interest.
For the local search part, once a solution is selected among the current best solutions,
a new solution for each bat is generated locally using random walk
xnew = xold + εAt (11)

where ε is a random number vector drawn from [–1, 1], while At = 〈 Ait 〉 is the average
loudness of all the bats at this time step.
Initially, each bat is randomly assigned a frequency which is drawn uniformly from
[fmin, fmax]. For this reason, bat algorithm can be considered as a frequency tuning
130 Ö. Tosun and M.K. Marichelvam

algorithm to provide a balanced combination of exploration and exploitation. The

loudness and pulse emission rates essentially provide a mechanism for automatic control
and auto zooming into the region with promising solutions.

3.4 Loudness and pulse emission

The loudness Ai and the rate of pulse emission (ri) have to be updated accordingly as the
iterations proceed. As the loudness usually decreases once a bat has found its prey, while
the rate of pulse emission increases, the loudness can be chosen as any value of
convenience. For simplicity, Ao = 1 and Amin = 0 can be used, assuming Amin = 0 means
that a bat has just found the prey and temporarily stop emitting any sound. So;
Ait +1 = α Ait , rit +1 = ri0 [1 − exp(−γt ) ] (12)

where α and γ are constants. In fact, α is similar to the cooling factor of a cooling
schedule in the simulated annealing. For any 0 < α < 1 and γ > 0,
Ait → 0, rit → ri0 , as t → ∞ (13)

In the simplest case, α = γ can be selected, and α = γ = 0.9 is used in this study.
The basic steps of the BA can be summarised as below:

Objective function f(x), x = (x1…xd)T

Initialise the bat population xi (i = 1, 2…n) and vi
Define pulse frequency fi at xi
Initialise pulse rates ri and the loudness Ai
while (t < Max number of iterations)
Generate new solutions by adjusting frequency, and updating
velocities and locations/solutions
if (rand > ri)
Select a solution among the best solutions
Generate a local solution around the selected best solution
end if
Generate a new solution by flying randomly
if (rand < Ai & f(xi) < f(x*))
Accept the new solutions
Increase ri and reduce Ai
end if
Rank the bats and find the current best
end while
Post process results and visualisation
 Hybrid bat algorithm for flow shop scheduling problems 131

4 Hybrid bat algorithm

Similar to many metaheuristic algorithms, in the bat algorithm the initial solutions are
generated randomly. But, studies have reported that the performance of the metaheuristic
algorithms might be improved by generating some of the initial solutions by using some
constructive heuristics that incorporate the knowledge of specific type of problems of
interest.
Ruiz and Maroto (2005) proved that the NEH heuristic is the best constructive
heuristic to minimise the makespan for the flow shop scheduling problems. Therefore, in
this paper it is planned to incorporate the NEH heuristic to solve the flow shop
scheduling problems. One of the initial solutions is obtained by the NEH heuristic is used
as an initial solution and the remaining solutions are generated randomly.
Preliminary studies by Yang (2010) suggested that BA is very promising for solving
nonlinear global optimisation problems. However, the major drawback of applying BA to
combinatorial problems is due to its continuous nature. In this HBA, the smallest position
value (SPV) rule described by Bean (1994) is introduced to convert the continuous
position in the bat algorithm to the discrete job permutation.
In the HBA for the flow shop scheduling problems, parameters were initialised and
N-1 population was generated randomly. One solution is generated using the NEH
heuristics. Each bat is initialised with some random position. The SPV rule applies to
each bat to find its corresponding permutation. Thus, each bat will be evaluated by using
the permutation to compute the objective function for the flow shop scheduling problems
with makespan criterion. Then the position is updated and new permutations are obtained.
The process is repeated until the stopping criterion is reached.

4.1 Solution representation

Solution representation is one of the most important issues in designing a HBA. The
solution search space consists of n dimensions as n number of jobs is considered in this
paper. Each dimension represents a job. The vector X t = ( X 1t , X 2t ,… , X nt ) represents the
continuous position values of bats in the search space. The SPV rule is used to
convert the continuous position values to the discrete job permutation. The solution
representation of a bat with four jobs is illustrated in Table 1.
Table 1 Example of SPV rule

Dimension j
1 2 3 4
Xij 0.68 0.32 0.16 0.44
Jobs 1 2 3 4

The SPV is 0.16 and the dimension j = 3 is assigned to be the first job in the permutation
according to the SPV rule. The second SPV is 0.32 and the dimension j = 2 is assigned to
be the second job in the permutation. Similarly, all the jobs are assigned in the
permutation. Hence, the sequence is 3-2-4-1.
132 Ö. Tosun and M.K. Marichelvam

5 Computational results

In this section, hybrid bat algorithm is tested with the different benchmark
problems of Reeves (1995). These datasets are obtained from the OR Library
(http://people.brunel.ac.uk/~mastjjb/jeb/info.html). There are 21 instances in Reeves
problems. This dataset has 20 to 75 jobs and 5 to 20 machines.
In the algorithm, population size is taken as 20 and the maximum number of the
iteration is limited to 500. For each test instance, 20 independent runs are carried out, the
best solution gained and the average of the 20 run is given. Pulse rate is selected from the
interval (0, 1) and frequency is from the interval (0, 150). The proposed HBA was coded
in C++ and run on a Pentium 4 3.4 GHz PC with 512 MB memory. For each problem
instances, the relative percentage increase from the best known solutions (BRE %) and
the average relative percentage error (ARE %) are also given. BRE value is calculated by
using the following formula:
f ABC − f min
% BRE = ⋅100 (14)
f min
In Table 2, algorithms used for comparison are given. Eight different algorithms are
chosen from the literature and results of the each algorithm are selected from these
studies. Performance of the algorithm for Reeves’ instances is given in Table 3,
comparing with the algorithms given in Table 2. BRE and ARE values of the each
algorithm is given.
Table 2 Metaheuristic methods used for the performance comparison with hybrid bat
algorithm
Study Proposed algorithm Properties
Tang et al. (2010) PGA GA integrated with PSO
Liu et al. (2007) PSOMA Hybrid PSO with local search and NEH
Liu et al. (2011) PSO-EDA_PI PSO modified with local search and
estimation of distribution algorithm
Rajkumar and IGA Improved GA incorporating with
Shahabudeen (2009) multi-crossover operators, multi-mutation
operators and hyper mutation
Wang and Zheng (2003) HGA Hybrid GA based on NEH, simulated
annealing and multi-crossover operators
Qian and Wang (2008) HDE Hybrid differential evolution
algorithm based on local search
and largest-order-value rule
Tosun (2014) ABC Hybrid artificial bee colony algorithm
with local search and NEH
Luo et al. (2014) DBA Discrete bat algorithm with NEH

For Reeves instances, hybrid bat algorithm has able to found best solutions of the nine
problem sets. If other metaheuristic methods are analysed, it is seen that ABC found 13,
PGA found 10, PSO-EDA_PI, HDE and IGA found 9, HGA and PSOMA found 8, finally
DBA found only 6. In general, for the 12 test instances proposed algorithm found the best
or better solutions from any of the solution methods given in Table 3. Only ABC has
more better (14) than the bat algorithm. But for the average BRE and ARE results, hybrid
bat algorithm yields the best values.
 Table 3
HBA NEH DBA ABC IGA HGA HDE PGA PSO_MA PSO-EDA_PI
BKS
BRE ARE RE BRE ARE BRE ARE BRE ARE BRE ARE BRE ARE BRE ARE BRE ARE BRE ARE
Rec01 1247 0 0.092 8.42 0 0.080 0 0.12 0 0.15 0 0.14 0 0.144 0 0.14 0 0.144 0 0.096
Rec03 1109 0 0 6.583 0 0.081 0 0.004 0 0 0 0.09 0 0 0 0.06 0 0.189 0 0.036
Rec05 1242 0 0.121 4.831 0.242 0.242 0 0.144 0 0.23 0 0.29 0.242 0.242 0.24 0.24 0.242 0.249 0.242 0.242
Rec07 1566 0 0.032 5.364 0 0.575 0 0.057 0 0.34 0 0.69 0 0.23 0 0.36 0 0.986 0 0
Rec09 1537 0 0 6.766 0 0.638 0 0 0 0.07 0 0.64 0 0 0 0.4 0 0.621 0 0.202
Rec11 1431 0 0 8.246 0 1.167 0 0 0 0.45 0 1.1 0 0 0 0.37 0 0.129 0 0.126
Rec13 1930 0.005 0.114 7.617 0.415 1.461 0 0.14 0.41 0.87 0.36 1.68 0.104 0.301 0 0.72 0.259 0.893 0.104 0.223
Rec15 1950 0 0.092 4.923 0.154 1.226 0 0.118 0 0.58 0.56 1.12 0 0.308 0 0.07 0.051 0.628 0 0.303
Rec17 1902 0 0.078 7.466 0.368 1.277 0 0.092 0 1.61 0.95 2.32 0 0.178 0 0.82 0 1.33 0 0.289
Rec19 2093 0.262 0.593 6.641 0.573 0.929 0.286 0.655 0.67 1.09 0.62 1.32 0.287 0.559 0.38 0.69 0.43 1.313 0.287 0.612
Rec21 2017 0.412 1.298 4.561 1.438 1.671 0.446 1.423 0.74 1.52 1.44 1.57 0.198 1.413 0.2 1.29 1.437 1.596 1.14 1.408
Rec23 2011 0.265 0.321 9.995 0.796 1.173 0 0.208 0.35 0.74 0.4 0.87 0.448 0.482 0.4 0.74 0.596 1.31 0.398 0.597
Rec25 2513 0.004 0.782 6.964 1.632 2.921 0 0.902 1.03 2.28 1.27 2.54 0.478 1.492 0.28 0.68 0.835 2.085 0.279 1.894
Rec27 2373 0.227 0.73 8.512 1.011 1.419 0.253 0.939 0.97 1.76 1.1 1.83 0.843 1.285 0.25 0.86 1.348 1.605 0.969 1.584
Rec29 2287 0.381 0.671 5.422 1.049 2.580 0.437 0.839 1.01 2.15 1.4 2.7 0.306 0.791 0.87 1.42 1.442 1.888 0.35 1.045
Rec31 3045 0.343 1.196 10.279 2.299 3.392 0.36 1.667 0.49 1.23 0.43 1.34 0.296 0.824 0.43 1.63 1.51 2.254 0.263 0.43
Rec33 3114 0.008 0.362 4.753 0.610 0.728 0 0.435 0.22 0.78 0 0.78 0 0.434 0 0.16 0 0.645 0 0.469
Hybrid bat algorithm for flow shop scheduling problems

Rec35 3277 0 0 5.005 0 0.037 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Rec37 4951 2.08 2.522 9.141 3.373 4.872 2.363 3.288 4.09 4.87 3.75 4.9 1.818 2.727 1.94 3.25 2.101 3.537 1.838 2.725
Rec39 5087 1.028 1.423 8.646 2.280 3.851 1.042 1.689 2.24 2.75 2.2 2.79 0.983 1.541 1.98 2.17 1.553 2.426 0.924 1.409
Rec41 4960 1.485 2.468 10.692 3.810 5.095 2.157 3.136 3.44 4.09 3.64 4.92 1.673 2.649 3.38 4.86 2.641 3.684 1.815 2.506
Performance evaluation of the hybrid bat algorithm for Reeves problems

Average 0.310 0.614 7.182 0.955 1.686 0.350 0.755 0.746 1.312 0.863 1.601 0.366 0.743 0.493 0.997 0.688 1.310 0.410 0.771
Note: Best solutions are given in bold.
133
134 Ö. Tosun and M.K. Marichelvam

5.1 Statistical analysis

In this section, statistical analysis of the proposed HBA algorithm for Reeves test
instances is given by using non-parametric tests.
Non-parametric tests can be used for comparing algorithms whose results represent
average values for each problem in spite of the inexistence of relationships among them.
Therefore these algorithms can be treated with probabilistic and non-probabilistic
methods without any restriction. Given that the non-parametric tests do not require
explicit conditions for being conducted, it is recommendable that the sample of results is
obtained following the same criterion, that is, computing the same aggregation (average,
mode, etc.) over the same number of runs for each algorithm and problem (Tosun, 2014;
Garcia et al., 2009; Luengo et al., 2009). All statistical analyses are done by using SPSS.
All the algorithms given in Table 2 except PSO-EDA_PI and DBA used 20 runs for
each problem. Therefore, PSO-EDA_PI and DBA are excluded from the statistical
analysis for a fair comparison.

5.1.1 Pairwise comparisons

Pairwise comparisons are the simplest type of statistical tests that a researcher can apply
within the framework of an experimental study. Such tests are directed to compare the
performance of two algorithms when applied to a common set of problems. In
multi-problem analysis, a value for each pair algorithm/problem is required (often an
average value from several runs). In this section, first a quick and easy, yet not very
powerful, procedure, which can provide a first snapshot about the comparison is used: the
Sign test. Then, Wilcoxon signed ranks test, as an example of a simple, yet safe and
robust, non-parametric test for pairwise statistical comparisons will be given (Deraac
et al., 2011).
Examples thorough this section will focus in characterising the behaviour of HBA
algorithm, in 1 × 1 comparisons with the rest of algorithms considered.
A popular way to compare the overall performances of algorithms is to count the
number of cases on which an algorithm is the overall winner, the number of wins is
distributed according to a binomial distribution (Derrac et al., 2011).
Table 4 shows the critical number of wins needed to achieve both α = 0.05 and
α = 0.10 levels of significance. In this study number of cases is 21.
Table 4 Critical values for the two-tailed Sign test at α = 0.05 and α = 0.10

# cases 5 6 7 8 9 10 11 12 13 14 15
α = 0.05 5 6 7 7 8 9 9 10 10 11 12
α = 0.10 5 6 6 7 7 8 9 9 10 10 11
# cases 16 17 18 19 20 21 22 23 24 25
α = 0.05 12 13 13 14 15 15 16 17 18 18
α = 0.10 12 12 13 13 14 14 15 16 16 17
Note: An algorithm is significantly better than another if it performs better on at least the
cases presented in each row.
 Hybrid bat algorithm for flow shop scheduling problems 135

For Reeves test instances, performing a Sign test to compare the results of HBA only
requires to count the number of wins achieved by HBA or by the comparison algorithm.
Then using Table 4, cases where a significant difference is detected can be highlighted.
This process is given in Table 5.
Table 5 Sign test results for pairwise comparison

HBA ABC IGA HGA HDE PGA PSOMA

Wins (+) 17 19 20 15 16 20
Loses (–) 1 - - 2 4 -
Detected differences α = 0.05 α = 0.05 α = 0.05 α = 0.05 α = 0.05 α = 0.05
Note: ABC shows a significant improvement over all the algorithms with a level of
significance α = 0.05.
The Wilcoxon signed rank test is used for answering the question “do two samples
represent two different populations?” It is a non-parametric test used in hypothesis testing
with a design of two samples. This is analogous to the paired t-test in non-parametric
statistical procedures; therefore, it is a pairwise test that aims to detect significant
differences between two sample means, the behaviour of two algorithms (Derrac et al.,
2011). Test results are given in Table 6.
Table 6 Wilcoxon sign test results

Comparison R+ R– p-value Comparison R+ R– p-value

HBA versus ABC 161 10 0.00 HBA versus HDE 137 16 0.00
HBA versus IGA 190 0 0.00 HBA versus PGA 192 18 0.00
HBA versus HGA 210 0 0.00 HBA versus PSOMA 210 0 0.00
Note: HBA shows an improvement over all the algorithms with a level of significance
α = 0.01.
According to the pairwise comparisons, proposed HBA algorithm shows improvements
over each of the other algorithms.

5.1.2 Friedman test

Friedman test (Friedman two-way analysis of variance by ranks) is a non-parametric test
equivalent of the two-way analysis of variance. Under the null hypothesis, it states that all
the algorithms are equivalent, so a rejection of this hypothesis implies the existence of
differences among the performance of all the algorithms studied. Friedman test ranks the
algorithms for each dataset separately, the best performing algorithm getting the rank of
1, the second best rank 2, and so on (Garcia et al., 2009; Derrac et al., 2011; Luengo
et al., 2009). Friedman test results are given in Table 7.
In Table 7, it is seen that HBA is the best performing algorithm for comparison with a
rank of 1.62 for the Friedman test. The p-value computed through the statistic of test
strongly suggests the significant differences among the algorithms compared.
136 Ö. Tosun and M.K. Marichelvam

Table 7 Ranks achieved by the Friedman test

Algorithms Friedman
HBA 1.62
ABC 2.95
IGA 5.00
HGA 6.33
HDE 2.88
PGA 3.62
PSOMA 5.60
Statistic 82.223
p-value 0.00

6 Conclusions

In this paper, the hybrid bat algorithm is applied to solve the PFSS problem under
makespan criterion. Initially developed for numerical optimisation, SPV rule is used to
modify the algorithm to use in flow shop problems. The performance of the algorithm is
tested by using the benchmark problems of Reeves. Algorithm has found the best
solutions for nine problems. Comparing with the other algorithms, proposed bat
algorithm has the lowest ARE and BRE values, which shows the quality of the proposed
approach. The proposed BA can also be implemented to solve the real industrial
scheduling problems. This algorithm would be useful in determining the optimal or near
optimal solutions and hence very much helpful to the managers.
For further studies to improve the performance of the bat algorithm, parameters
optimisation may be carried out. Different local search techniques may be used to
improve the performance of the proposed algorithm. Consideration of sequence
dependent setup time and transportation time are some other future scope of this research
work.
Proposed algorithm can further be developed with other local search techniques and
their performance can be compared. Also by using different techniques for converting
continuous solutions positions to discrete, performance of the algorithm can easily
effected. This HBA algorithm will also be used for other scheduling problems (like job
shop scheduling or hybrid flow shop scheduling) or different engineering optimisation
problems with some little adaptation.

Acknowledgements

This paper was supported by The Scientific Research Projects Coordination Unit of
Akdeniz University. The authors would like to thank the Editor-in-Chief of the journal of
IJMOR, Prof. Angappa Gunasekaran, and the anonymous reviewers for their helpful
comments and valuable suggestions that improved the quality of this paper.
 Hybrid bat algorithm for flow shop scheduling problems 137

References
Bean, J.C. (1994) ‘Genetic algorithms and random keys for sequencing and optimization’,
INFORMS Journal on Computing, Vol. 6, No. 2, pp.154–160.
Campbell, H.G., Dudek, R.A. and Smith, M.L. (1970) ‘A heuristic algorithm for the n job, m
machine sequencing problem’, Management Science, Vol. 16, No. 10, pp.630–637.
Derrac, J., Garcia, S., Molina, D. and Herrera, F. (2011) ‘A practical tutorial on the use of
nonparametric statistical tests as a methodology for comparing evolutionary and swarm
intelligence algorithms’, Swarm Evolutionary Computing, Vol. 1, No. 1, pp.3–18.
Framinan, J.M., Gupta, J.N.D. and Leisten, R. (2004) ‘A review and classification of heuristics for
permutation flow-shop scheduling with makespan objective’, Journal of Operations Research
Society, Vol. 55, No. 12, pp.1243–1255.
Framinan, J.M., Leisten, R. and Ruiz-Usano, R. (2005) ‘Comparison of heuristics for flowtime
minimization in permutation flowshops’, Computers and Operations Research, Vol. 32, No. 5,
pp.1237–1254.
Gandomi, A.H. and Yang, X.S. (2014) ‘Chaotic bat algorithm’, Journal of Computational Science,
Vol. 5, No. 2, pp.224–232.
Garcia, S., Molina, D., Lozano, M. and Herrera, F. (2009) ‘A study on the use of non-parametric
tests for analyzing the evolutionary algorithms’ behavior: a case study on the CEC’2005
Special Session on Real Parameter Optimization’, Journal of Heuristics, Vol. 15, No. 6,
pp.617–644.
Johnson, S.M. (1954) ‘Optimal two and three-stage production schedules with set-up times
included’, Naval Research Logistics Quarterly, Vol. 1, No. 1, pp.61–68.
Liu, B., Wang, L. and Jin, Y.H. (2007) ‘An effective PSO-based memetic algorithm for flow shop
scheduling’, IEEE Transactions on System, Man and Cybernetics Part B, Vol. 37, pp.18–27.
Liu, H., Gao, L. and Pan, Q. (2011) ‘A hybrid particle swarm optimization with estimation of
distribution algorithm for solving permutation flowshop scheduling problem’, Expert Systems
with Applications, Vol. 38, No. 4, pp.4348–4360.
Luengo, J., Garcia, S. and Herrera, F. (2009) ‘A study on the use of statistical tests
for experimentation with neural networks: analysis of parametric test conditions and
non-parametric test’, Expert Systems with Applications, Vol. 36, No. 4, pp.7798–7808.
Luo, Q., Zhou, Y.Q., Xie, J., Ma, M.Z. and Li, L.L. (2014) ‘Discrete bat algorithm for optimal
problem of permutation flow shop scheduling’, The Scientific World Journal, dx.doi.org/10.
1155/2014/630280.
Marichelvam, M.K. (2012) ‘An improved hybrid cuckoo search metaheuristics algorithm for
permutation flow shop scheduling problems’, International Journal of Bio-Inspired
Computations, Vol. 4, No. 4, pp.200–205.
Marichelvam, M.K., Prabaharan, T., Yang, X.S. and Geetha, M. (2013) ‘Solving hybrid flow shop
scheduling problems using bat algorithm’, International Journal of Logistics Economics and
Globalisation, Vol. 5, No. 1, pp.15–29.
Mittal, M.L. and Kanda, A. (2009) ‘Scheduling of multiple projects with resource transfers’,
International Journal of Mathematics in Operational Research, Vol. 1, No. 3, pp.303–325.
Najid, N.M. and Arroub, M. (2010) ‘An efficient algorithm for the multi-mode resource
constrained project scheduling problem with resource flexibility’, International Journal of
Mathematics in Operational Research, Vol. 2, No. 6, pp.748–761.
Nawaz, M., Enscore, E. and Ham, I. (1983) ‘A heuristic algorithm for the m- machine, n- machine
flow shop sequencing problem’, Omega, Vol. 11, No. 1, pp.91–95.
Qian, B. and Wang, L. (2008) ‘A hybrid differential evolution method for permutation flow-shop
scheduling’, International Journal of Advanced Manufacturing Technologies, Vol. 38,
Nos. 7–8, pp.757–777.
138 Ö. Tosun and M.K. Marichelvam

Rajkumar, R. and Shahabudeen, P. (2009) ‘An improved genetic algorithm for the flowshop
scheduling problem’, International Journal of Production Research, Vol. 47, No. 1,
pp.233–249.
Reeves, C.R. (1995) ‘A genetic algorithm for flowshop sequencing’, Computers and Operations
Research, Vol. 22, No. 1, pp.5–13.
Ruiz, R. and Maroto, C. (2005) ‘A comprehensive review and evaluation of permutation flowshop
heuristics’, European Journal of Operational Research, Vol. 165, No. 2, pp.479–494.
Sabba, S. and Chikhi, S. (2014) ‘A discrete binary version of bat algorithm for multidimensional
knapsack problem’, International Journal of Bio-Inspired Computations, Vol. 6, No. 2,
pp.140–152.
Shang, J. and McKendall, A.R. (2009) ‘Solution techniques for a crane sequencing problem’,
International Journal of Mathematics in Operational Research, Vol. 1, No. 4, pp.433–455.
Sharma, S., Gupta, D. and Sharma, S. (2013) ‘Bicriteria scheduling on parallel machines: total
tardiness and weighted flowtime in fuzzy environment’, International Journal of Mathematics
in Operational Research, Vol. 5, No. 4, pp.492–507.
Tang, J., Zhang, G., Lin, B. and Zhang, B. (2010) ‘Hybrid genetic algorithm for flow shop
scheduling problem’, International Conference on Intelligent Computation Technology and
Automation, Vol. 2, pp.449–452.
Tosun, Ö. (2014) ‘Using artificial bee colony algorithm for permutation flow shop scheduling
problem under makespan criterion’, International Journal of Mathematical Modelling and
Numerical Optimisation, Vol. 5, No. 3, pp.191–209.
Wang, H.Y., Lu, Y.B. and Peng, W.L. (2013) ‘Permutation flow-shop scheduling using a hybrid
differential evolution algorithm’, International Journal of Computing Science and
Mathematics, Vol. 4, No. 3, pp.298–307.
Wang, L. and Zheng, D.Z. (2003) ‘An effective hybrid heuristic for flow shop scheduling’,
International Journal of Advanced Manufacturing Technologies, Vol. 21, No. 1, pp.38–44.
Yang, X.S. (2010) ‘Firefly algorithm, stochastic test functions and design optimisation’,
International Journal of Bio-Inspired Computations, Vol. 2, No. 2, pp.78–84.
Yang, X.S. (2011) ‘Bat algorithm for multi-objective optimisation’, International Journal of
Bio-Inspired Computation, Vol. 3, No. 5, pp.267–274.
Yang, X.S. (2013) ‘Bat algorithm: literature review and applications’, International Journal of
Bio-Inspired Computations, Vol. 5, No. 3, pp.141–149.
Ying, K.C. (2012) ‘Minimising makespan for multistage hybrid flowshop scheduling problems
with multiprocessor tasks by a hybrid immune algorithm’, European Journal of Industrial
Engineering, Vol. 6, No. 2, pp.199–215.
Zobolas, G.I., Tarantilis, C.D. and Ioannou, G. (2009) ‘Minimizing makespan in permutation flow
shop scheduling problems using a hybrid metaheuristic algorithm’, Computers and Operations
Research, Vol. 36, No. 4, pp.1249–1267.