Investigating Optimum Resourse Allocation in University

Course Timetabling using Tabu Search: An Incremental

Strategy

Mehtap KOSE ULUKOK1, Yasemin ERCAN2

1,2
Cyprus International University, Computer Engineering Department, Haspolat, KKTC
1
[email protected], [email protected]

Abstract. In this study, the possible solutions of the Tabu Search algorithm is
investigated for the solution of the University Course Timetabling Problem
(UCTP). The diffuculty of the UCTP is the coordination of lectures, classrooms
and teachers according to some constraints. The replacement of courses and
lectures to periods and classrooms is the main task that have to be performed for
each academic semester. This replacement requires new solution for each
semester because of part-time lecturers, changes on workloads of lecturers and the
changes in the number of students. Therefore, the UCTP must be solved every
semester again and again. The main aim of this study is to solve the UCTP with
the use of optimum number of classroom assignment by reducing the unused
periods in classrooms’ timetable. Tabu Search incremental strategy is studied
within the Faculty of Engineering at Cyprus International University (CIU).

Key words: timetabling problem, tabu search algorithm, metaheuristics

1 Introduction
The UCTP has been choosen by many researchers since 1965. It is a NP-hard problem
and it is a real-world application of combinatorial optimization problems. Timetabling
problems are mainly worked in high schools and universities. Usually, these problems
are solved manually. Depending on the search space of the problem these manual
solutions often require a few weeks or more to solve, because during the construction
process of timetabling always some changes are done. After some number of
modifications, the result is mostly not optimal. Over the past few years, researchers
have been investigated new algorithms or techniques to find general solution of
timetabling problem. However, the UCTP solutions may vary university to university
because of their own curriculum and therefore they require indivual solutions.
Timetabling problems are studied with several techniques for more than four decade.
Some of the recent studies includes Constraint-Based Reasoning [1], Integer
Programming [2], Tabu Search Algorithm [3], Memetic Algorithms [4], Genetic
Algorithms [5]-[7-9]. Other popular techniques including Graph Coloring Algorithms
[6] and Linear Algorithms [10]. Acceptable solutions were reported in use of all these
techniques, but most of the time these solutions require man touch.

A mathematical model of university timetables is introduced as a linear program [10].

The timetabling problem is a type of multi-objective satisfaction problem. Mainly the
constraints are classified into two as hard and soft constraints. The vaoilation of these
constraints denotes the quality of the developed solution. Therefore, the solution space
of the problem is huge and finding the optimal solution is a complex task. Studies show
that Genetic Algorithms find optimal solutions [5],[7]-[9] but they are complicated with
respect to time and memory. Morever, recently a computer tool is developed to solve
the UCTP with tabu search algorithm and efficient solutions are reported [3]. Another
computer tool is developed for the generalized solution of UCTP at Yeditepe University
called TEDI [7].

2 Tabu Search Algorithm for UCTP

Tabu search is one of the commonly used metaheuristic algorithm for solving
combinatorial optimization problems, such as traveling salesman problem (TSP),
university course timetabling problem (UCTP), etc. It provides optimal or near optimal
solutions for the practical problems. Tabu search algorithm searches the solution
domain by exploiting the neighbors of the current solution using a tabu list. Some of the
basic ingredients of the tabu search are a tabu list, neigborhood and aspiraiton criteria.

A tabu list keeps the list of not allowed and visited neighbors during the search process.
The main aim of tabu list is to avoid reversal of moves and cycling. Moves are
performed between adjacent solutions that can be reached from the current solution. If a
tabu move results in better solution than any previously performed move so far, than
this tabu move may be ignored. A condition that allows such a move is called an
aspiration criteria.

The timetabling problem constraints are mainly considered as hard constraints in which
every timetable must satisfy all these constraints, and soft constraints which are not
essential for a timetable solution but it makes the solution better when they solved.

Almost in all of the past studies similar hard and soft consraints are listed [1]-[3].
Depending on the university’s limitations these constraints may vary. In this study, the
number of classroom assignment is aimed to minimize to reduce the university resource
usage cost.

In this study, a novel tabu search algorithm is proposed to solve the timetabling problem
with incremental strategy. The proposed algorithm starts with a randomly generated
solution to the UCTP and this initial solution is improved by using incremental strategy
where a best movement is choosen in a classroom time table by satisfying the hard
constraints. Possible moves of tabu search algorithm is peprformed for all courses of
the active department incrementally. If there is any hard constraint violation, this
solution gets worst fitness values. All periods in a classroom timetable is searched first
to fullfill the classroom periods. If there is no suitable period available, then a new
classroom period is started to search.

2.1 Parameters and Constraints of UCTP

The course timetabling problem deals with finding the exact time slots of courses within
a limited periods, classrooms and teachers. The problem parameters are given in Table
1. Courses in each department usually organized in number of semesters. Constraints
regarding to the semesters of each courses and about teachers are organized as hard
constraints and soft constraints. Hard constraints are the ones that must be obeyed. Soft
constraints are those constraints that make the solution better but they are not fully
essential. The used hard constraints are:

 No classroom can be assigned to more than one course at the same timeslot.
 No teacher can be assigned to more than one course at the same timeslot.
 Same semester courses cannot be assigned to the same timeslot.

On the other hand used soft contraints are:

 Two hours gap between teachers’ lectures.

 Don’t assign lecture within the teacher’s reserved timeslots.

Table 1. CIU, Faculty of Engineering timetabling problem parameters

Parameters Value
Number of courses 108
Number of rooms 16
Number of teachers 40
Number of days 5
Number of periods 9

The proposed tabu search algorithm is illustrated in Fig. 1. Algorithm starts with a
randomly generated solution and seraches the solution space for better solution based on
hard and soft constraints with a tabu list. Moves are incrementally choosen from the
assigned timeslots. Two types of moves are used in the proposed tabu search algorithm;
simple move and simple swap which are illustrated in Fig. 2 and Fig. 3, respectively.
 Start

Gerenate a random initial solution

Find candidate list of solutions

Update
the tabu Evaluate quality of solutions
list

Choose the best solution

Does the stoping condition meet?

Stop

Fig. 1. Tabu search algorithm: Incremental strategy

The simple move shifts the choosen course timeslot into another empty slot if there
exits any in current room or in one of the other used rooms.

Periods Mon Tue Wed Thu Fri

09:00 MAT101
10:00 MAT101
11:00 MAT101
12:00 MAT101
13:00
14:00
15:00
16:00
17:00

Fig. 2. Simple move

However, the simple swap swaps two courses blok of timeslots where they are assigned.
Moves of the proposed tabu search algorithm are marked as tabu once they have been
performed. This leads to protect correct timeslot assignments and to avoid recycling.

Periods Mon Tue Wed Thu Fri

09:00 ENG101
10:00 ENG101
11:00
12:00
13:00
14:00
15:00 CPE112
16:00 CPE112
17:00

Fig. 3. Swap move

2.3 Results of Tabu Search: Incremental Strategy

The proposed algorithm starts to solve the timetabling problem in the Faculty of
Engineering department by department. Mainly four departments are activelly offering
courses to the faculty and these departments’ courses are scheduled according to
departments’ pre-defined order. This is aimed to reduce the time and memory
requirements of the tabu search algorithm and it is also aimed to minimize the search
space of the problem domain. The performance of the proposed algorithm is
summarized in Table 2. The results show that the proposed algorithm acheives good
performance with respect to timeslots utilization and the used number of classrooms
according to the given parameters in Table 1.

Table 2. Analyze of man-made solution of CIU course timetabling with the produced tabu search
solution for the Faculty of Engineering.

Man-made Tabu Search

Objective
solution produced solution
Violation of hard constraints 0 0
Violation of soft constraints 6 8
Number of used rooms 16 12
Overall timeslots utilization 70,63% 80%

In order to compare an algorithm with other studies, all most similar parameters must be
used. The number of courses that are teaching by instructors and the nunber of part-time
instructors effect the hard constraints of the UCTP. However, when the timeslots
utilizations are considered in [1] and [3], the performance of these stduies are given as
80% and 95% respectively. In [1], constraint-based reasoning strategy is investigated
with 16 rooms and 136 courses for UCTP for the University of Yemen, but there is no
information about the number of teachers. Another tabu search algorithm for the
solution of UCTP is given in [3] having 57 courses, 6 rooms and 27 teachers. The
workload of the teachers in Statistics Department of Hacettepe University seems that 2
on the average. However, the proposed tabu search algorithm applied on 3 different
courses workload for each teacher on the average. This is because, the proposed tabu
search algorithm has lowest timeslot utilization than [3].

The developed software GUI screen shot is shown in Fig. 4. The implementation of tabu
search incremental strategy has been performed in C# programming language by using
a laptop which has Intel Core 2 Duo 2.2GHz processor with 4 GB RAM.

Fig. 4. An example screen shot of a found solution for UCTP

5 Conclusıon
The proposed tabu search algorithm is investigated for the solution of timetabling
problem in Faculty of Engineering at Cyprus International University (CIU). The
developed computer software will be used in all faculties of CIU. CIU has same first
two years lectures in the all departmetns of the Faculty of Engineering and it consisting
of seven departments. Last two years of each department differs from each other.
Thereby, the whole problem search space is reduced to consecutive two year curriculum
courses.
One of the aim of this study to develop a computer automated software which solves
UCTP for the CIU with optimum classroom usage. A user friendly computer graphical
user interface will be developed to solve other faculties timetabling problems in CIU.

The results of the proposed algorithm are compared with the man made solution of CIU
Faculty of Engineering and also with studies in other universities. The solution of the
proposed tabu search algorithm for the UCTP is encouraging with means of optimum
resourse allocation and timeslots utilization.

References
1. Irene, H.S.F., Safaai-Deris, Hashim, S.Z. (2009). Investigating Constraint-Based
Reasoning for University Timetabling Problem. In Proceedings of the International
MultiConference of Engineers and Computer Scientists. Hong Kong, 18 – 20
March. IMECS 2009.

2. Bakır, M.A., and Aksop, C. (2008). A 0-1 Integer Programming Approach to a

University Timetabling Problem. Hacettepe Journal of Mathematics and Statistics.
37(1): 41-55.

3. Aladağ, Ç.A., and Hocaoğlu, G. (2007). A Tabu Search Algorithm to Solve a

Course Timetabling Problem. Hacettepe Journal of Mathematics and Statistics. 36
(1): 53-64.

4. Özcan, E. and Alkan, A. (2007). A Memetic Algorithm for Solving a Timetabling

Problem: An Incremental Strategy. In Proceedings of the 3rd Multidisciplinary Int.
Conf. On Scheduling: Theory and Application. Paris, France. 28-31 August. P.
Baptiste, G. Kendall, A. M. Kordon, F. Sourd (ed.)

5. Nuntasen, N. and Innet, S. (2007). A Novel Approach of Genetic Algorithm for

Solving University Timetabling Problems: a case study of Thai Universities. In
Proceedings of the 7th WSEAS International Conference on Applied Computer
Science. Venice, Italy, 21-23 November.

6. Ülker, Ö., Özcan, E., Korkmaz, E.E. (2006). Linear Linkage Encoding in Grouping
Problems: Applications on Graph Coloring and Timetabling. In Proceedings of the
6th International Conference on the Practice and Theory of Automated
Timetabling. Brno, Czech Republic, 30 August- 1 September. LNCS 3867,
Springer.

7. Özcan, E. And Alkan, A. (2002). Timetabling using a Steady State Genetic

Algorithm. In Proceedings of the 4th International Conference on the Practice and
Theory of Automated Timetabling. Gent, Belgium, 21-23 August. LNCS 2740,
Springer.

8. Burke, E., Elliman, D., Weare, R. (1994). A Genetic Based University Timetabling
System. In Proceedings of the 2nd East-West International Conference on
Computer Technologies in Education.Crime, Ukraine, September. EW-ED’94.
9. Colorni, A., Dorigo, M., Maniezzo, V. (1990). Genetic Algorithms and Highly
Constrained Problems: The Time-Table. In Proceedings of the 1st International
Workshop on Parallel Problem Solving from Nature. Dortmund, Germany. LNCS
496, Springer.

10. Akkoyunlu, E.A. (1994). A Linear Algorithm for Computing the Optimum
University Timetable. Computational Journal. 16 (4): 347-350.

Biographies
Mehtap KÖSE ULUKÖK- Born in Silivri at 1975. Graduated from Computer
Engineering Department of Eastern Mediterrenean University, Famagusta, T.R.N.C at
1998. Master and PhD programs were completed in Computer Engineering Department
from the same university at September 2000 and January 2009, respectively. Artificial
intelligence is the main research topic.

She has several publications about metaheuristic searches, expert systems and robot
control.

Yasemin ERCAN- Born in Gaziantep at 1978. Graduated from Computer Studies in

Information Technology in Eastern Mediterrenean University, Famagusta, T.R.N.C at
2001. Master program was completed from Information Systems Department of Faculty
of Fine Arts in the same university at June 2003. She is currently studiying in
Management Information System Department of Cyprus International University for her
PhD degree. Internet programming and database systems are the main research topics.

She has some publications about expert systems, web programming with database
systems and tabu search algorithm.