Using Integer Programming to Solve the Crew Scheduling Problem in

the Taipei Rapid Transit Corporation

Chui-Yen, Chen
1,2

1
Department of Industrial Engineering and Management
University of Yuan-Ze, Ph.D. student
2
Department of Tourism and Leisure Management
Chin-Min Institute of Technology, Instructor
110, Hsue-Fu Rd., Tou-Fen, Miao-Li 351, Taiwan, R.O.C.
Tel: 886-37-605670 Fax: 886-37-605674
[email protected]


ABSTRACT: The transport operation volume has reached millions of people since the Taipei Rapid Transit
Corporation (TRTC) started operating 28 March 1996. TRTC has become the key public transportation in
Taipei city. Crew scheduling of station employees has become one of the basic costs of operating the Public
Rapid Transit System. During the planning stage of the project, a mathematical model of planning phases was
established and used to find answers to crew scheduling and reduce costs. In general, the crew scheduling
problems relate to the set covering problem or set partitioning problem; therefore, it is often described by
the methods of 0-1 Integer Programming. Hence, this study will adopt the methods of integer programming for
mathematical programming to establish a crew scheduling program. The Target-based formula of using integer
programming to solve the crew shift scheduling problem is meant to meet the interests of station employees.
Furthermore, this study will use comparison analysis to research (1) common shift scheduling; (2) employees
preferred working hours; and (3) specified work day for employees to help solve day off or other problems. We
hope to provide this method to the TRTC in a rapid and convenient timetable to improve the efficiency of
operations and business competition.

Key-words: Crew scheduling problem (CSP), 0-1 Integer programming, Taipei Rapid Transit Corporation
(TRTC)


1. Introduction
Human resource management is one of the major
issues for services industries. Recently, awakening
labor interests have included law enforcement of
limitations on working hours and labors efforts to
establish a union to protect laborers. These and other
limitations, therefore, push us to use the shortage of
human resources to arrange an optimistic shifting
schedule. This section will discuss the Taipei Rapid
Transit Corporation (TRTC) station employees crew
scheduling problem.
Since the TRTC starting operating on 28 March
1996, the transport operation volume has reached
millions of people. Therefore, the TRTC has become
the key public transportation in Taipei, Taiwan.
Because the TRTC is a year-round operation and it
offers long service hours, staff working regulations
differ from those of the common business employee.
Currently, staff schedules for TRTC are divided into
three shifts. In addition, the shifting schedules must
comply with requirements set by the Labor Standards
Law during each scheduled shift. Traditionally, shift
scheduling is completed manually. This manual shift
scheduling not only causes inefficiencies in
employee operations, but also wastes time and costs.
We expect that we can establish a convenient and
rapid timetable via this research institute.
Because Integer Planning is categorized into
NP-Completeness, the model variations may be
excessive; however, it might address the issue of
efficiency. Although it might be difficult to find an
optimistic solution to crew scheduling, this study
proposes that we can build a model to use as a trial.
This model would test crew scheduling for the most
employee stations in the TRTC of the Taipei main
station (Ban-Nan line). For example, to search for the
largest interest among station employees, this study
might solve the model via LINGO.



WSEAS TRANSACTIONS on
INFORMATION SCIENCE & APPLICATIONS

Chui-Yen, Chen
ISSN: 1790-0832
331
Issue 4, Volume 5, April 2008
2. Literature Review
The crew shift scheduling problem is the most
common issue in mathematical science. Related
studies are too numerous to enumerate, and they vary
widely. For the crew shift scheduling problem,
whether it be crew shift scheduling or crew
allotment, both are related to multi-specifications and
become an issue for 0-1 round number or complex
Integer Planning. In addition, Bartholdi [12]
suggested that the crew shift scheduling problem is
related to NP-Completes problem. Therefore, we are
focused on large-scale problems and find it difficult
to discover an optimum solution under reasonable
algorithms. We always adopt a heuristic method
solution to find the most efficiency and nearly
optimum solution within a short time. Table 1
summarizes the major literature on this issue.
Table 1. Literature review of crew shift scheduling
Author Subjects Methods Key issue
Browmell and
Lowerre [2]
Cyclic scheduling
Method of
mathematics
Uses the method of mathematics to performthe periodic shift scheduling
operations. Used to probe into how the shift scheduling policy affects
human resources requirements.
Chuang, Shuo
ping [3]
Service employee Fuzzy theory
Uses the concept of triangular fuzzy function to select the evaluation
criterion, and also arrange existing evaluation methods to analyze the
adoption and advantages or disadvantages.
Uses the methods of AHP giving weight to build the evaluation construction
of unity quantification, and induce the clustering analysis as the decision of
group framework, under fuzzy number.
Su, Chao-Ming,
Chang, Ching [10]
Taipei Rapid Transit
System(TRTS)'s
station employee
Mathematical
programming
Establishes two stages of shift scheduling models: the first stage determines
the day off and working day for each station employee; the second stage
decides the working hours for the station employee. Through the two stages,
the authors build a model and use LINDO software to solve the problem.
Kuo, Wei -Yin
[8]
Registered nurse Integer planning
Uses the methods of mathematical programming to perform the shift
scheduling programfor Registered Nurses and builds the shift scheduling
model, as well as using the Linear Programming Software Package LINDO
to find a solution.
Wang, Pao-Chih
[11]
Tank truck driver
Mathematical
programming
decision support
system(DSS)
This study suggests that the decision support system(DSS) is under the goal
of outsourcing un-transportable mileage and drivers un-transportable
mirage. By comparison with a Chinese petroleum company (CPC) in
Taiwan, the original shift scheduling has improved from66.2% to 77.3%;
thus the effect is excellent.
Huang, J ung-Hua,
Tsai, Chih- Hao
[5]
Paramedic
Mathematical
programming
This study first focuses on the employer's problemof paramedics in hospital
and developing algorithms, then under the conditions of cost-saving and
complementing human resources to search for an optimumemployee hiring
range. After that, the study focuses on the crew shift scheduling problem,
establishing the model of mathematical programming. In order to improve
the goal of work-satisfaction for the paramedic, and also under the condition
of regulations, the operations policies and a variety of environmental
requirements are used to search for optimumoff day and in-shift timetable.
For the application of software solution, this study used the LINDO as the
crucial part.
Liu, Shuang- Huo
[9]
Passenger
Transportation
Service Staff for
Taiwan railway
Combined
Genetic
Algorithmwith
An Ant Colony
Optimization
System
While you account for the condition for train schedule/manpower factors
and limitations, you can adopt the genetic algorithms to solve between the
human workload balance and minimumoverall work hours problemfor
train transportation. In doing so, you build the Passenger Transportation
allocation model, and then combine the shift scheduling within each shifts
problem. Finally, you use the Ant Colony Algorithms to search for the
optimumsolution in the shifts model of total shift periods.
Huang, J en- Yu
[4]
Timetable for MRT
Ant Colony
Optimization
Algorithms
Uses the methods of Ant Colony Optimization as its basis, and uses the
characteristic of Ant Colony Optimization Algorithms easily running the
parallel-process. After combined with the parallel-process technology, you
can use the Taipei MRT Route Map and related stations, combining this
information with time and convert it into the internet pattern. At the same
time, you take the characteristics of off-peak/rush hours for train scheduling,
conflict time adjustment as well as the variety of depots drive in-out hour
into account, in order to build a combined basic shift scheduling with
conflict adjustment over two stages shift produce algorithms.
This study is based on the industrys crew shift
scheduling differences, thus producing different
solutions. Beasely and Cao [1] suggested that the
problem of crew shift scheduling lies within three
categories by industry: 1) airliners crew shift
scheduling; 2) public transportation service
employees shift scheduling; and 3) official
employees shift scheduling. Because the public
transportation service employees crew shift
scheduling is similar to airliners crew shift
scheduling, portions of this study used the steps to
solving airliner crew shift scheduling as the solving
process. The TRTCs shift scheduling is a
smaller-scale problem than airliners, due to official
employees shift scheduling problem and the in-shift
hours or rest regulations, which are mostly known or

WSEAS TRANSACTIONS on
INFORMATION SCIENCE & APPLICATIONS
Chui-Yen, Chen
ISSN: 1790-0832
332
Issue 4, Volume 5, April 2008
fixed. When compared with the massive level of
airliner crew shift scheduling problems, the time to
find the questions solution is smaller and more
unified. The complexity of the problem is also lower
than that for the airline industry. Therefore, some
studies adopted LINDO or LINGO directly to solve
the problem.
Because the crew scheduling literature review is
arranged by years, we can identify trends for crew
shift scheduling and also understand some of the
previous/current methods of crew scheduling. In
early studies, crew scheduling almost always uses the
fixed pattern, such as periodic. With variations,
solutions gradually become multi-property, primarily
for the purpose of enhancing the employees affect
factor (such as fairness, reasonability, employees
anticipation). For the shift scheduling pattern, we
take the in-shifts and days-off problem into account.
Furthermore, the computer has innovated rapidly and
has become universal for arranging shift scheduling
or auxiliary shift scheduling. Thus, the computer has
increased the convenience and the efficiency of
scheduling and has become one of the important
tools for the universe of business.
Through the scholars suggested methods of
shift scheduling procedures, we can understand the
implementation of the shift scheduling procedures
under various situations. In early studies, the scholar
always uses mathematical methods to perform
periodic shift scheduling procedures; thus the
consideration factors and the methods are easier.
With increasing tools and methods, later scholars
have coordinated mathematical programming
methods with applying Computer Programming
Languages or heuristic algorithms, for example. The
consideration factors are embedded not only in the
consideration of periodic shift scheduling, but also in
the increased number of shift systems for shift
scheduling performance for specific occupations or
contractors (such as the nurse, full-time teacher
school scheduling, 104 operators shift scheduling in
Taiwan, and stewardess, etc). For each different
careers requirements, we take special situations into
account in order to obtain more accurate and efficient
shift scheduling.


3. Model Construction and Evaluation
We first follow the limitation condition, subject and
hypothesis, to build the station employees shift
scheduling model. We then convert the collected data
into the analysis practice, and investigate the
applications of this model. We hereby use the
following symbols and follow the alphabetical order
to define the meaning, as below:
i Station employee number,
Deputy Station Master: ,station employee: .
n m i ... ... 1 =
m n
j In shifts, p j ... 1 =
k Work days
j
d In shifts employee requirement
UB
i
t The maxima limitation working days of i
th

station employee
LB
i
t The minimum limitation working days of i
th

station employee
ij
w The station employees preference for in
shift hour or day
ij
x {0, 1} Decision variables
ik
y The working shift is 0 or 1 of i
th
station
employee on day k

This study is completed knowing daily shifting
conditions to create the shift scheduling for TRTCs
station employees. The Target-based formula is to
meet the largest interest of station employees, as
follows:

= =
m
i
p
j
ij ij
x w Max
1 1

(1)
subject to
j d x
j
n
i
ij
=

=1
(2)
i t x t
UB
i
p
j
ij
LB
i

=1

(3)
6 ... 1 6
20
=

+
=
p q i x
q
q j
ij

(4)
2 ... 1 1
2
=

+
=
p q i x
q
q j
ij

(5)
m i j x
m
i
ij
... 1 1
1
=

=

(6)
,... 9 , 6 , 3 0 = = j female i x
ij

(7)
j i x
ij
,
1
0

=

(8)
The target formula (1) is focused on the
preference weight of station employees in shifts. In
summary, the goal of this section is to figure out the
maximum value.
The limit formula above (2), is the on-duty
employee number=the requirement value of

WSEAS TRANSACTIONS on
INFORMATION SCIENCE & APPLICATIONS
Chui-Yen, Chen
ISSN: 1790-0832
333
Issue 4, Volume 5, April 2008
on-dutys station employee. Formula (3) is the
maximum/minimum limitations of a station
employees working hour, and also the condition for
any station employees special working conditions.
Formula (4) is the continuous working days (no more
than six days). After the employee has worked
continuously for six days they should have one day
off. Formula (5) is when the station employee
performs the shift; in the case of continuous working
for two-three days, they must take two shifts off. For
example, when the employee works on a day shift,
he can work on any shift the next day. If he works on
the Swing shift, he can only work on the Swing shift
the next day. Formula (6) is for complying with the
regulation that each employee in shift must have a
supervisor or deputy station master. Formula (7) is
for the female employees shift scheduling
regulations. Formula (8) is the 0-1 limitation of
decision variables .
ij
x

= =
+

m
i
n
k
ik ik
y y
1 1
1
(9)
For the evaluation model, this study focuses on
evaluating the differences among station employees
continuous work situations, such as formula (9). If
the station employee works continuously for two
days (
1
1
=
+ ik
y
and
1 =
ik
y
) or took two days off
(
0
1
=
+ ik
y
and
0 =
ik
y
), then formula (9) is zero (0).
If the station employee works continuously for one
day and took one day off (the and value, one
is zero (0), the other value is one (1)),
the
1 + ik
y
ik
y
ik ik
y y
+1
value is one (1), as is shown in Table
2, Example 1, the value is three (3)
(=|1-1|+|0-1|+|1-0|+|1-1|+|0-1|). In Table 1, Example
2, the value is one (1)
1(=|1-1|+|1-1|+|1-1|+|0-1|+|0-0|). The lower the value,
the difference for continuous working shift is lower,
therefore, it implies that each work shift is closer.

Table 2. the description of shift scheduling models evaluation
Example In shifts scheduling
ik ik
y y
+1

I DDRDDR 3
II EEEERR 1
Note: We used four of six working days for this example, with D being the day shift, E the Swing shift, and R is a day off.

The Intel Pentium D 3.4GHz CPU, the memory
is DDR-II 800 2,048MB, hard disk is 250GB,
7200rpm, and the processing software is
mathematical programming software: LINGO 8.0
was used to perform the testing of this study models.


4. Scenario Analysis and Testing
Results
The example of this study is TRTCs station
employee. It uses the requirement of a daily in-shift
employee, under known conditions. It focuses on the
shift scheduling planning of the station employee,
and investigates whether the station employees
assignment is a problem. For theoretical analysis, we
used the most crowded operating station of the
TRTC, the Taipei main station (Ban-Nan line) as an
example and to perform the testing (see Figure 1).
For the employee requirements of the
TRTC-Taipei main station (Ban-Nan line), the
stations employee requirement is current the largest.
The stations in-shift employee requirement is now
17 shift, the female employee is 8 shifts, and deputy
station master is 5s. In every station, however, we
only deploy a Station Master, whose on-duty hour is
fixed. The in-shift time is 08:30-17:30; he doesnt
in-shift with other station employees. Hence, this
study does not include this factor. In addition, the
in-shift scheduling is three shifts: the day shift time
is 06:30~14:30 (D); swing shift time is: 14:15~22:15
(E) and night shift time is: 22:00~06:45(N). The
on-duty employees required for the day shift and
swing shift is 5s daily, and we also deploy a deputy
station master on every on-duty shift. The night shift
requires two employees, one is the deputy station
master, and the other is a station service employee.
To estimate manpower, TRTC focuses on the
hardware facility, and does not concern itself with
the number of passengers; therefore, there is no issue
with rush/off peak commuting hours. After the
employee finishes an on-duty shift, he must take a
break of two shifts. The employees continuous
working days cannot exceed more than six days;
therefore, they must take one day off after working
continuously for six days. In addition, by regulation
of Labor Standards Law in Taiwan, the employee
must take eight days off in one month. If they have
an extra day off, their off days will exceed eight
days. We also respect female employees rights, so
the female employee of the TRTC is only on duty
WSEAS TRANSACTIONS on
INFORMATION SCIENCE & APPLICATIONS
Chui-Yen, Chen
ISSN: 1790-0832
334
Issue 4, Volume 5, April 2008
during the day shift and swing shift. This study used
the example of the monthly shift hours; the shift
scheduling monthly is 90 shifts.


Fig.1 Taipei MRT Route Map

To account for the interests of the station
employee, this study asked the station employees and
deputy station masters to fill out a questionnaire,
weighing their interests monthly. The maximum
weight of an interest in a working day is (50), with
the minimum weight of a less interesting working
day is (10). To avoid the station employee or deputy
station master filling out the same shift of the interest
working day, and also result in some shift where no
one is on duty, this study has setup the maxima
weight into 10, but 0, meaning that the station
employee still has chance to work on a special shift,
and therefore, it would lower the requirements of
employee. In addition, the night shift of female
station employees weight is zero(0),
0 =
ij
x
,the
reason for doing this is due to we must meet the
limitation formula (7), which regulates that female
employee do not work the night shift.
Therefore, under normal model operations of
shift scheduling, we can distinguish if it with the
interest of station employees testing or not. Model A
shows the station employees interest as common;
meaning that the station employee doesnt have the
right to choose working days and in-shift schedules.
Model B is based on the interest of station employee
by adding up the weight factors, where the station
employee can choose his/her interest in shift
scheduling. Because one of the goals is to satisfy of
the preferences of station employee, model B allows
them to search for the best interest to work with the
shift scheduling. Model C is to test for the
employees most interest in a working day. The three
kinds of shift scheduling models, A, B and C, are
shown in Table 3.

WSEAS TRANSACTIONS on
INFORMATION SCIENCE & APPLICATIONS
Chui-Yen, Chen
ISSN: 1790-0832
335
Issue 4, Volume 5, April 2008


Table 3. The shift scheduling arrangement of model A
Employee
number
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Day off(days) 8 9 9 8 9 8 9 8 9 8 9 9 9 10 9 11 8
Day shift 8 8 6 9 6 8 12 10 15 13 12 8 2 8 8 11 6
Swing shift 5 6 7 6 15 14 9 12 6 9 9 13 9 4 9 5 12
Night shift 9 7 8 7 0 0 0 0 0 0 0 0 10 8 4 3 4

Table 4. The shift scheduling arrangement of model B
Employee
number
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Day off(days) 8 9 8 8 8 9 9 8 10 8 10 9 9 8 10 9 10
Day shift 6 8 5 7 13 11 6 14 7 13 10 13 8 8 9 9 3
Swing shift 3 6 8 9 9 10 15 8 13 9 10 8 9 7 7 7 12
Night shift 13 7 9 6 0 0 0 0 0 0 0 0 4 7 4 5 5

Table 5. The shift scheduling arrangement of model C
Employee
number
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Day off(days) 8 9 8 8 8 9 8 10 9 8 9 9 10 8 11 9 9
Day shift 5 6 5 9 12 14 6 11 8 16 12 6 8 13 9 4 6
Swing shift 3 12 8 5 10 7 16 9 13 6 9 15 8 6 4 10 9
Night shift 14 3 9 8 0 0 0 0 0 0 0 0 4 3 6 7 6

After collecting the data above, and arranging it
into the table we arrived at the following: the
employee number, 1-5 are deputy station masters,
and the employee number, 5-12 are female
employees. They do not work during the night shift.
Generally, the station employees days off amounted
to more than 8-11 days, so it would comply with the
Labor Standards Law in Taiwan. By comparing the
shift scheduling attachment, you can see there is a
deputy station master on duty between each shift.
Therefore, the three-models all comply with the shift
scheduling of TRTC.
If you need to further investigate which model
is optimum, you can use this study models
evaluation index to investigate the working
frequency of each station employee. As Table 6
shows, we set the evaluation value of Model C into
less, because this model is better than both Model A
and Model B. In addition, by comparing the
frequency of continuous days off, Model A and
Model B both focus on continuous two days off. For
the Model C, because the employee can fill the
schedule taking into account personal working
interests, the frequency of two or three continuous
two days off is more even.


Table 6. The comparison of each model
Evaluation subject Model A Model B Model C
Model evaluation value 200
*
220
*
186
*
The sum up of continuous two days off 20 21 13
The sum up of continuous three days off 10 7 16
The sum up of continuous four days off 1 0 1
Solution Optimal Optimal Feasible
Object Value 3,600 15,170 15,630(15,640)
**
Time 9 sec 11sec 7min 07sec
Iterations 11,788 18,174 456,380
*
the models evaluation value is the summary for 17s employee
**
The target value is 15,640, but we can find the Feasible Solution on target value 15,630.

WSEAS TRANSACTIONS on
INFORMATION SCIENCE & APPLICATIONS
Chui-Yen, Chen
ISSN: 1790-0832
336
Issue 4, Volume 5, April 2008

In addition, if you work on the model process
procedure of LINGO, then Model A and Model B
both all with the solution methods. You can with
Feasible Solution only on Model C. For process time
consumption, Model C took longer than the others.
In contrast to Models A and B, Model Cs process
frequency was higher than the others because it was
restricted by the working interest day value of the
employee.


5. Conclusion and Suggestions
This studys conclusions and suggestions are:

1. This study established a crew shift
scheduling model of TRTC through
mathematical programming, and used the
LINGO to search for the applicable
solution model. The result show how this
model is applied for the current crew shift
scheduling of TRTCs employee.
2. Currently, the TRTCs crew shift
scheduling process is done manually, so
rapidly changing the crew shift scheduling
is inefficient. This study is done through
the mathematical programming software
LINDO for the solution to this issue.
3. For complying with manpower
requirements, station employees working
days and days off, this model does not
only provide the common shift scheduling
for TRTC, but also deals with the interests
of employees for working day and days off
for through preference weighting, in order
to solve the actual situation of shift
scheduling.
4. In addition, TRTCs day-off regulations
must comply with Labor Standards Law,
and female employees do not work the
night shift. You can use the restrictions of
this model to find the solution.
5. Using the evaluation models index, you
can obtain crew shift scheduling from each
model, station employees working
frequency. Then you can use the working
frequency's figure to determine equitable
shift scheduling. Using the case study, you
can see that the result of Model C is better
than others.
6. For Model C, the target of this models
value is 15,640. While using LINGO, you
could only find the Feasible Solution of
target value is 15,630. This section
suggested that you could use the method
of mathematical programming or heuristic
method algorithms to resolve this.
7. In the case study, this study only focuses
on the most needed employees of TRTC
operation station, currently (Taipei main
station [Ban-Nan line]). While the
variation could not use LINGO to find the
solution, you could use the heuristic
method algorithms to find the solution for
TRTC station employees shift scheduling
problem. We expected that the
requirement of solution process frequency
would be lower than the current crew shift
scheduling, and could also improve the
efficiency of shift scheduling.
8. Currently, the structure of this study can
only arrange the shift scheduling for a
single station. We hope to extend this
model to the whole shift scheduling in the
future in order to comply with the shift
scheduling process for TRTC and improve
the efficiency of crew shift scheduling.


REFERENCES
[1] Beasely, J .E and Cao, B. (1996). A tree
Search Algorithm for the Crew Scheduling
Problem, European Journal of
Operational Research, 94(3), pp. 517-526.
[2] Browmell, W. S. and Lowerre, J . S. (1976).
"Scheduling of Work Force Required in
Continuous Operations under Alternative
Labor Policies," Management Science,
22(5), pp597-605.
[3] Chuang, Shuo ping. (1997). The
Evaluation of service employees
timetabling problem. The Graduate of
Business Management:Yuan-Ze
University.
[4] Huang J en-Yu. (2005). Using Ant Colony
Optimization to solve the TRTCs train
scheduling problem. The Graduate of
Technology Management:Chung-Hua
University.
[5] Huang J ung-Hua and Tsai Chih-Hao.
(2003) An Integrated Model of Nurse
Staffing and Scheduling, Management
Review of Fu-Jen Catholic University,
10(3), pp.105-138.
[6] Hung, R., (1999). Scheduling a
Workforce under Annualized Hours,
International Journal of Production
Research, 37(11), pp.2419-2427.

WSEAS TRANSACTIONS on
INFORMATION SCIENCE & APPLICATIONS
Chui-Yen, Chen
ISSN: 1790-0832
337
Issue 4, Volume 5, April 2008
[7] J acobs, L.W. and Loucks, J .S. (1991).
Tour Scheduling and Task Assignment of
a Heterogeneous Work Force: A Heuristic
Approach, Decision Science, 22,
pp.719-737.
[8] Kuo Wei-Yin. (2001). Designing a
Nursing Scheduling for Hospital-An
Application of Integer Goal
Programming. The Graduate of Health
Services Administration:China Medical
University.
[9] Liu Shuang-Huo. (2005). Solving the
TRAs Conductor Scheduling Problem by
Genetic Algorithms and Ant Colony
Optimization, The Graduate of Business
Management:Ming-Hsin University.
[10] Su, Chao- Ming and Chang, Ching. (2000).
Solving the Crew Scheduling Problem by
a two-stage 0-1 Integer Programming
Model, Journal of Transportation, 12(2),
pp.1-14.
[11] Wang, Pao-Chih. (2002). A prototype of
decision support system for scheduling
drivers of tank trucks of the Chinese
Petroleum Corp. The Graduate of
Technology Management:Chung-Hua
University.
[12] Yan, Shangyao and Yu-Ping, Tu (2002).
A Network Model for Airline Cabin
Crew Scheduling, European Journal of
Operational Research, 140, pp.531-540.


WSEAS TRANSACTIONS on
INFORMATION SCIENCE & APPLICATIONS
Chui-Yen, Chen
ISSN: 1790-0832
338
Issue 4, Volume 5, April 2008
Appendix
Table A.1 Model A: the crew shift schedule without any employees performance
NO. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1 MON N R D D E R D E D E E R N R E R D
2 TUS N R D E R E D E D R E D N E R R D
3 WEN N E D N D E E E D D R E R R R D R
4 THUR N E E R D E R E D D R E R N R D D
5 FRI R N R D E E R R D D R E D N E D E
6 SAT E N R D R R E D D D D E E R N R E
7 SUN R N R D E E E D R D D R E D N R E
8 MON D N D R E E E R D R D R E D N R E
9 TUS D R N R E E E R D D R R E D N D E
10 WEN D E N R E E R R D D D D E N R E R
11 THUR D R N E E R D D E E D R R N D E R
12 FRI D D R N E D E E R E D D R N R R E
13 SAT E D D N R D R E R E R D N R D E E
14 SUN R D E N R E D E E R D E N D D R R
15 MON R D E N R E D R E E D E R R D D N
16 TUS N E R R D E D D R R D E E R E D N
17 WEN N E D D R R D D E R E E N R E D R
18 THUR N R R E D R D D E E E R N D E R D
19 FRI R R E N D E E R R E E D N D R D D
20 SAT E N E R D R R D D E R E N E R D D
21 SUN R N E D R D R D D E R E R E E D N
22 MON E N R D E D R D D R E E D N R E R
23 TUS N R R D E D R D R D E E E N D R E
24 WEN N R E D E R D E D D E R N R D R E
25 THUR R D N E E E D R D R R R N D D E E
26 FRI D D N R E E D R D D R E R R E N E
27 SAT E D N R R R D E E D D R E D E N R
28 SUN R D N E E D R E R D D D R E E N R
29 MON D R N E E D E E R D D D R N R R E
30 TUS D E R N R D E E R R E D E R D D N


WSEAS TRANSACTIONS on
INFORMATION SCIENCE & APPLICATIONS
Chui-Yen, Chen
ISSN: 1790-0832
339
Issue 4, Volume 5, April 2008

Table A.2 Model B: the crew shift schedule with employees prefer hours
NO. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1 MON R R E N D R E D D E E D E R R D N
2 TUS D R E N D E E D E R E D N R R D R
3 WEN D N N R E E E D R D R D R D R E E
4 THUR E N N D R R R D D D R E D E E R E
5 FRI N N R E D D R E R D E R D E R D E
6 SAT N R R E D D E R D E R D E R D N E
7 SUN N D E R E E R D E R D R E D D R N
8 MON R E N D E R D R E D E R E D D R N
9 TUS D R N R E D E R E D R E N D D E R
10 WEN E D N R R D R E E D D E R N D E R
11 THUR N D R E D D D E R E D E R R E N R
12 FRI N D D E D E D E R R E E R D R R N
13 SAT R N D E R E E E R D E R N D D D R
14 SUN R R E N D R E R E D R D N E D E D
15 MON R D N R E D R D E E D E R R D N E
16 TUS N D R D E E D E R R D R D R E N E
17 WEN N D D E R R D R D R E D E E E R N
18 THUR R R E N D R E D D E E D E N R D R
19 FRI D R E N D E E D E R R D R N R D E
20 SAT D E N R E R E D R D R D D N R E E
21 SUN E E N D R D R D D E R E D N E R R
22 MON N E R D D E R E E R E R D N R D D
23 TUS N N R E D R E R E D R D E R E D D
24 WEN R N E R D D R D E D D R E E N R E
25 THUR D N R D E D D D E E E R R N R R E
26 FRI N R D R E E E D R E R D D R D N E
27 SAT N E N D R E E R D R D E D D E R R
28 SUN N E R E D R E E E D D R R D N D R
29 MON N R D E R D E R R D D D R E N E E
30 TUS R D E N R D R D R E D D E E N E R


WSEAS TRANSACTIONS on
INFORMATION SCIENCE & APPLICATIONS
Chui-Yen, Chen
ISSN: 1790-0832
340
Issue 4, Volume 5, April 2008

Table A.2 Model C: the crew shift schedule with employees prefer days
NO. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1 MON E R N D D R D D E R E E N R R D E
2 TUS N R R D E R D D E D R E N D R E E
3 WEN N E D D E D E D R D R E R E R N R
4 THUR N R E D R D R E D D R E D E E N R
5 FRI N D R E D E R E D E E R D N R R D
6 SAT R E R N D E E R D E R E D N D R D
7 SUN N E E R D E R E D R D R E N D R D
8 MON R N R E D R D R E D E E N R D E D
9 TUS D N D R E D E R E R E R N R D E D
10 WEN E N D R E R E R R D E D R D D E N
11 THUR N R E D R D E D E D E R R D E N R
12 FRI N R E D E D E D R D E D R E R R N
13 SAT N D E E R D E D R E R D R E D N R
14 SUN R D E N R E R D R E R D D E D N E
15 MON R E R N D E E D D E D R D R N R E
16 TUS R E N N D E E R D R D D E R R D E
17 WEN E E N R D R R D E R D D E R D E N
18 THUR N R R D E R D D E R D E E D R E N
19 FRI N R D N E R D E E D D E R D R E R
20 SAT N R D R E D E E R D E E D R R N D
21 SUN N D E D R D R E D R R E R D E N E
22 MON R D N E R R D E E D E R D D N R E
23 TUS N E R N D D E R E D R E D D R R E
24 WEN R E E N D D R E R D D R E E D R N
25 THUR D E R N D D E R D E D R E R E R N
26 FRI D E N R D E E R E R D R E D N D R
27 SAT D R N R E R E R E D D E E D N D R
28 SUN N D N R E D E D E D R E R D R E R
29 MON R E N D R D E E R D D E R D N E R
30 TUS D E N E R D R R R D D E R D N E E



WSEAS TRANSACTIONS on
INFORMATION SCIENCE & APPLICATIONS
Chui-Yen, Chen
ISSN: 1790-0832
341
Issue 4, Volume 5, April 2008