Mathematical Modelling of Engineering Problems

Vol. 8, No. 3, June, 2021, pp. 377-385

Journal homepage: http://iieta.org/journals/mmep

Mathematical Model and Efficiency of Courses Timetable Creation Process

Oleksii Yuriy Sakaliuk*, Fedir Anatoly Trishyn

Department of Technological Processes Automation and Robotic Systems, Odessa National Academy of Food Technologies,
Odessa 65039, Ukraine

Corresponding Author Email: [email protected]

https://doi.org/10.18280/mmep.080306 ABSTRACT

Received: 6 April 2021 The object of the study is the courses timetable creation process. A complex of problems
Accepted: 18 May 2021 has been solved, which are related to the mathematical description of the control object.
The main purpose of the paper is development of a complex of mathematical models,
Keywords: which are describing the courses timetable creation process and a set of criteria for
automation, automated control system, assessing the quality of a class schedule in higher education. The observation method was
control object, identification, optimality used researching of the courses timetable creation process. The research of the process was
criterion, Pareto compromise, scheduling conducted on the basis of the timetable department of the Odessa National Academy of
Food Technologies using the information system "Rozklad". The observer was included
in the process creation of the courses timetabling as a participant. Observation is controlled
by degree of formalization and conducted in real conditions. Regularity is not systematic
namely participants in the process (researcher and staff) deal with unplanned phenomena
and unexpected situations. The method of formalization and analysis was used to describe
the mathematical model of the courses timetable creation process. The result of the paper
is the obtained mathematical models of constraints that are imposed on the timetable and
a description of the efficiency criteria of the resulting schedule. The prospects for further
research may include the development of the logical control system for the courses
timetable creation process, experimental research and optimization of the proposed criteria
with fulfilling all the proposed conditions for conducting the process.

1. INTRODUCTION of this study, the assignment of classes to teachers, and then

the distribution of these classes over time is an important
The task of scheduling classes is an integral part of the work administrative task that must be performed each semester. An
of any educational institution, including higher education important feature of this schedule is that it can be adjusted
institutions. From the point of view of formalization in the during the semester and the curriculum for the semester must
theory of the schedule, the schedule of classes is the be followed.
determination on the time scale of the place of the discipline There is a wide variety of works describing a wide range of
with the fulfilment of the requirements set for them. In order educational methods of scheduling. A review of its literature
to develop an effective automation control system of the is contained in ref. [2]. The variety of methods and approaches
courses timetable creation process, it is necessary to define all to solving the problem of planning suggests that there is still
the requirements and connections between them [1]. One of no universal solution. Common to all researchers is the finding
the steps in developing a process control automation system is of different heuristics that reduce the number of complete
a mathematical description of the control object. The need for search operations. Each of the researchers offers his approach
research is to determine the impact of parameters and to solving the problem of scheduling. Examples of such
limitations of the process of the courses timetable creation methods are mathematical and integer programming,
process. That is, obtaining a mathematical model that constraint-based methods, the Hungarian method, methods
describes the properties and characteristics of the control based on graph theory, genetic algorithm, TABU-search,
object. neural networks and evolutionary methods.
The relevance of the work for automation and planning is We will note that courses timetable creation process
due to the fact that most of the work on this subject are depends on the result of the decision of other tasks which need
outdated, or are incomplete reviews of the tasks of creating an to be solved at the organization of the educational process. The
effective schedule of classes and automation of scheduling, organization of the educational process is based on the Laws
they address only certain aspects of the problem. of Ukraine "On Education", "On Higher Education", "On
For more than half a century, the time schedule problems Scientific and Scientific-Technical Activity", normative legal
have attracted the attention of scientists around the world. One acts of the Verkhovna Rada of Ukraine, the Cabinet of
of the typical problems of planning is the courses timetable Ministers of Ukraine, the Ministry of Education and Science,
creation. The problem of scheduling is to schedule a set of the National Quality Assurance Agency higher education,
meetings between teachers and students over a period of time standards of educational activity and standards of higher
that requires certain resources and meets some additional education and other normative documents [1].
requirements. In the university schedule, which is the subject The formation of mandatory and desirable requirements and

377
determining the importance of these requirements is the initial participant in the educational process (teacher), student [9].
stage of formalizing the task of forming a schedule of classes. Lessons are formed for each discipline and student group.
A lot of scientists have been engaged in this task [3-6]. The The set of lessons can be represented as a combination of two
main mandatory requirements, non-compliance with which subsets: 𝐿 = 𝐿 𝑇𝑆𝑇 ∪ 𝐿𝐹𝑆𝑇 .
will lead to non-compliance with the curriculum, were LTST– the set of lessons with special requirements for
described. Researchers have described the desired audience, 𝐿 𝑇𝑆𝑇 = {𝐿 𝑇𝑆𝑇 𝑖 , 𝑖 = 1, 𝑁𝑙𝑡𝑠𝑡 } , where Nltst – the
requirements, but these requirements are not universal for all number of lessons with special requirements for audiences.
educational institutions. The main hard and soft constraints are LFST – the set of lessons without special requirements for
described. Lytvynenko et al. [7] provides a mathematical
audience, 𝐿𝐹𝑆𝑇 = {𝐿𝐹𝑆𝑇 𝑖 , 𝑖 = 1, 𝑁𝑙𝑓𝑠𝑡 } , where Nlfst – the
model of the classes scheduling problem in higher education
and considers the algorithm for solving the problem. In the number of lessons without special requirements for audiences.
work of A.S. Hasuhadzhiev [8] formed a set of requirements Let's enter a set of time slots 𝑇 = {𝑇𝑖 , 𝑖 = 1, 𝑁𝑡 }– the time
for the selection of an acceptable version of the schedule and intervals when lessons take place, where Nt – the total number
selected estimates of the importance of indicators for each of time intervals of the timetable. Usually, the schedule of
group of requirements. lessons is made for one semester. The number of weeks, days
The object of study is the courses timetable creation process. and pairs per day is known. Each element of the set of time
𝑖 𝑖 𝑖
The courses timetable creation process is a labor-intensive, slots is a triple type of tuple: 𝑇𝑖 = 𝑇(𝑇𝑤𝑒𝑒𝑘 , 𝑇𝑑𝑎𝑦 , 𝑇𝑝𝑎𝑖𝑟 ), where
tedious process that requires a significant amount of human 𝑖
𝑇𝑤𝑒𝑒𝑘 = 1, 𝑁𝑤𝑒𝑒𝑘 – the number of the week, 𝑇𝑑𝑎𝑦 𝑖
= {𝑇𝑖 , 𝑖 =
resources and time. Therefore, to increase the creation 𝑖
1, 𝑁𝑑𝑎𝑦 }– the number of the day, where 𝑇𝑝𝑎𝑖𝑟 = 1, 𝑁𝑝𝑎𝑖𝑟 – the
timetable speed and the quality of the developed schedule, you
need to automate the process. The first stage of automation is pair number during the day. 𝑁𝑤𝑒𝑒𝑘 , 𝑁𝑑𝑎𝑦 , 𝑁𝑝𝑎𝑖𝑟 – the number
the identification and description of mathematical models. of weeks, days and pairs in one semester [10].
The subject of study is a mathematical model of the courses D– the set of disciplines𝐷 = {𝐷𝑖 , 𝑖 = 1, 𝑁𝑑 }, where Nd – the
timetable creation process: a mathematical description of number of disciplines. The discipline can be represented as a
process variables, soft and hard constraints, efficiency criteria function of parameters: name of the discipline 𝐷𝑛𝑖 and lesson
of the schedule and its target function. duration in hours 𝐷𝑡𝑖 : 𝐷𝑖 = 𝐷(𝐷𝑛𝑖 , 𝐷𝑡𝑖 ).
The purpose of the work is to describe the mathematical
A – the set of audiences 𝐴 = {𝐴𝑖 , 𝑖 = 1, 𝑁𝑎 }, where Na – the
model and description of the efficiency criteria of the resulting
number of audiences of the educational institution. The set of
schedule.
the audiences can be represented as a combination of two
Now we give a formalized statement of the courses
subsets: 𝐴 = 𝐴 𝑇𝑆𝑇 ∪ 𝐴𝐹𝑆𝑇 .
timetable creation problem, the main condition of which is to
implement the curriculum. It is necessary to develop such a
mathematical model of courses timetable creation process,
which will describe the schedule with all hard constraints, as
well as minimize non-compliance with all soft constraints,
having only data on the audiences, the list of disciplines taught
in higher education, the list of participants educational process
(teachers) and students, and set time intervals in which classes
can take place. In addition to the courses timetable creation, it
is necessary to provide for the possibility of prompt adjustment
of the schedule in real-time.
The following task follows from the above. Figure 1. The hierarchical structure of the set “Audiences”
Let the given sets that are ordered by index:
-audiences, 𝐴 = {𝐴𝑖 , 𝑖 = 1, 𝑁𝑎 }; ATST – the set of specialized audiences, 𝐴 𝑇𝑆𝑇 = {𝐴 𝑇𝑆𝑇 𝑖 , 𝑖 =
-disciplines, 𝐷 = {𝐷𝑖 , 𝑖 = 1, 𝑁𝑑 }; 1, 𝑁𝑎𝑡𝑠𝑡 }, where Natst – the number of specialized audiences.
-teachers, 𝑃 = {𝑃𝑖 , 𝑖 = 1, 𝑁𝑝 }; Specialized audiences are the set of laboratories and audiences
-students, 𝑆 = {𝑆𝑖 , 𝑖 = 1, 𝑁𝑠 }; with special equipment 𝐴 𝑇𝑆𝑇𝑙 = {𝐴 𝑇𝑆𝑇𝑙 𝑖 , 𝑖 = 1, 𝑁𝑎𝑡𝑠𝑡𝑙 }, where
-timeslots, 𝑇 = {𝑇𝑖 , 𝑖 = 1, 𝑁𝑡 }. Natstl – the number of laboratories and audiences with special
equipment.
It is necessary to find such a set 𝐿 = {𝐿𝑖 , 𝑖 = 1, 𝑁𝑙 }, for all AFST – the set of non-dedicated audiences, 𝐴𝐹𝑆𝑇 =
items Li, where: 𝐹 = ∑𝐾 𝑖=1 𝜆𝑖 ⋅ 𝜙𝑖 → 𝑚𝑎𝑥. {𝐴𝐹𝑆𝑇 𝑖 , 𝑖 = 1, 𝑁𝑎𝑓𝑠𝑡 } , where Nafst – the number of non-
Moreover, the schedule is subject to hard constraints,
violation of which is unacceptable when creating the schedule dedicated audiences. Non-dedicated audiences are the set of
of courses. large audiences for students streams 𝐴𝐹𝑆𝑇𝑙 = {𝐴𝐹𝑆𝑇𝑙 𝑖 , 𝑖 =
1, 𝑁𝑎𝑓𝑠𝑡𝑙 }, where Nafstl – the number of large audiences for
students streams; the set medium audiences for
2. DETERMINING PROCESS VARIABLES groups 𝐴𝐹𝑆𝑇𝑚 = {𝐴𝐹𝑆𝑇𝑚 𝑖 , 𝑖 = 1, 𝑁𝑎𝑓𝑠𝑡𝑚 } , where Nafstm – the
number of medium audiences for groups; the set of small
Let's build a mathematical model of courses timetabling
audiences for subgroups 𝐴𝐹𝑆𝑇𝑠 = {𝐴𝐹𝑆𝑇𝑠 𝑖 , 𝑖 = 1, 𝑁𝑎𝑓𝑠𝑡𝑠 } ,
creation process, identifying variables and constraints.
where Nafsts – the number of small audiences for subgroups.
L– the set of lessons, 𝐿 = {𝐿𝑖 , 𝑖 = 1, 𝑁𝑙 }, where Nl – the
Figure 1 shows the hierarchical structure of audiences set.
number of lessons. Each of them is characterized by the Non-specialized audiences are audiences without special
following parameters: academic discipline, audience, equipment, such as lecture halls. Specialized audiences

378
include laboratories in which laboratory and practical work 𝐿𝑃𝑝 – the set of lessons, which involved teacher Pp. 𝐿𝑇𝑡 – the
takes place using special equipment, devices or individual set of lessons that take place in the time interval Tt. Thus, in
conditions. the equation (1) the amount 𝐿𝑖𝑝 from the set 𝐿𝑃𝑝 ∩ 𝐿𝑇𝑡 not more
The audiences can be represented as a function of than 1. This means that during this pair, the teacher conducts
parameters: audience volume 𝐴𝑖𝑣 , audience type 𝐴𝑖𝑡 = {0; 1}, this lesson, or he is free.
if𝐴𝑖𝑡 = 1, than the audience is specialized, otherwise without 2) No more than one lesson can be held in one audience
special purpose and the educational building in which the at a time:
audience is located 𝐴𝑖𝑏 : 𝐴𝑖 = 𝐴(𝐴𝑖𝑤 , 𝐴𝑖𝑡 , 𝐴𝑖𝑏 ).
P – the set of participants in the educational process 𝑃 = ( Aa , Tt ) : Aa  A; Tt  T ;
{𝑃𝑖 , 𝑖 = 1, 𝑁𝑝 }, where Np – the number of participants in the (!L : (A = A )  (L = L ))
i i a i
Tt
(2)
 (! L : ( A = A )  (L = L ))
educational process.
Tt
S – the set of students, 𝑆 = {𝑆𝑖 , 𝑖 = 1, 𝑁𝑠 }, where Ns – the i i a i
number of students. Students united in Nsg subgroups, forming
a set of subgroups Sg, which looks like: 𝑆𝑔 = {𝑆𝑔𝑖 , 𝑖 = 𝐿𝑇𝑡 – the set of lessons that take place in the time interval Tt.
1, 𝑁𝑠𝑔 } , where Nsg – the number of subgroups. Subgroups Expression (2) means that there is a single cycle of lessons,
which is assigned a given audience and time interval, or such
united in N g groups, forming a set of groups G, which looks a cycle does not exist and the audience is free.
like: 𝐺 = {𝐺𝑖 , 𝑖 = 1, 𝑁𝑔 }, where Ng – the number of groups. 3) No more than one lesson can be held per student at a
N g groups united in Nts students streams, where Nts – the time:

number of students streams. The set of students streams looks (S s , Tt ) : S s  S ; Tt  T ;

like: 𝑇𝑠 = {𝑇𝑠𝑖 , 𝑖 = 1, 𝑁𝑡𝑠 }.
Thus, a specific lesson can be represented as a function of (!L : (S = S ) (L = L ))
i
*
i s i
Tt
(3)
the parameters of the discipline, time slot, participant in the  (! L : (S = S ) (L = L ))
i
*
i s i
Tt
educational process and the student: 𝐿𝑖 =
𝐿(𝐷𝑑𝑖 , 𝑇𝑡𝑖 , 𝑃𝑝𝑖 , 𝑆𝑠𝑖 , 𝐴𝑖𝑎 , 𝐿𝑖𝑝 , 𝐿𝑖𝑠 , 𝐿𝑖𝑘𝑖𝑛𝑑 ) where 𝐷𝑑 ∈ 𝐷, 𝑇𝑡 ∈ 𝑇, 𝑃𝑝 ∈
𝐿𝑇𝑡 – the set of lessons that take place in the time interval Tt.
𝑃, 𝑆𝑠 ∈ 𝑆, 𝐴𝑎 ∈ 𝐴, 𝐿𝑖𝑘𝑖𝑛𝑑 – type of lesson and binary variables: The expression (3) means that there is a single cycle of lessons
𝐿𝑖𝑝 = {0; 1}, 𝐿𝑖𝑠 = {0; 1}. The meaning of these variables is as where a subset of students (student formation) are on the
follows. If 𝐿𝑖𝑝 = 1, then the teacher P in the time interval T lesson, or such a cycle does not exist and student formation is
must be in the audience A. At 𝐿𝑖𝑝 = 0 this statement is free.
incorrect. If 𝐿𝑖𝑠 = 1, then it means that the student S in the time 4) The number of students in the audience must not
interval T must be in the audience A. At 𝐿𝑖𝑠 = 0 this statement exceed the volume of the audience:
is incorrect.
( Aa , S s , Tt ) : Aa  A; S s  S ; Tt  T ;

3. HARD CONSTRAINTS DESCRIPTION  S si  Ava (4)

iL Aa  LS s  LTt

All the requirements for the courses timetable will be

divided into two groups: mandatory (hard parameters) and 𝐿𝑆𝑠 – the set of lessons, which are held for set of the students
desirable (soft parameters). Ss. 𝐿𝐴𝑎 – the set of lessons, which are held in the audience Aa.
As part of the courses timetabling creation, we highlight the LTt – the set of lessons that take place in the time interval Tt.
task of forming an acceptable timetable (not necessarily Thus, in the Eq. (1) the amount 𝑆𝑠𝑖 from the set 𝐿𝐴𝑎 ∩ 𝐿𝑆𝑠 ∩ 𝐿𝑇𝑡
optimal), which satisfies all the mandatory requirements. no more than the volume of the relevant audience 𝐴𝑎𝑣 . This
Failure to comply with any of the hard constraints means that means that in this time interval the number of students does
the courses timetable is not working. This means that there are not exceed the volume of the audience, or the audience is
no situations in the schedule when one teacher can conduct empty.
more than one lesson at a time, one audience holds more than
one lesson at a time, one student visits more than one audience
at a time, the number of students in the audience exceeds the 4. SOFT CONSTRAINTS DESCRIPTION
volume of the audience.
The task of courses timetabling creation is to determine the The task of courses timetabling creation puts forward
time interval for each audience, taking into account the increased demands on machine resources. Solving the problem
implementation of hard objective constraints and to provide by a complete search and the ratio of all options for the
the required number of cases of simultaneous student schedule of courses is almost impossible. In addition to the
formation (students groups, groups, subgroups) and the task of creation an acceptable schedule, we highlight the task
teacher in one of the possible audiences (laboratories) [7]. of improving the existing schedule by fulfilling the desired
1) One teacher can conduct no more than one lesson at requirements [11].
a time: 1) Lack of "windows" (this is a term used by the
supervisors of the educational department, which means a free
( )
 Pp , Tt : Pp  P; Tt  T ; time interval between classes is longer than a break [1]) for
students. Fewer "windows" in the schedule will improve the
 Lip  1 (1)
quality of the schedule, because classes will be placed more
Pp
i L  LTt densely and the student will not have to wait for the next class

379
more than a break time: ( )
 S s , Ttday : S s  S ; Ttday  T Tday ; ( )
( )
 S s , Ttday : S s  S ; Ttday  T Tday ; ( )
iLlectu res  L t
 Lis( T T day )
2
(10)

 Lis = (imax − imin ) + 1 (5)

iLS s  L t
(
T Tday )
Llectures – the set of classes such as "lecture". 𝐿𝑇𝑡(𝑇𝑑𝑎𝑦) – the
set of lessons that take place in the time interval Tt(Tday). Thus,
𝐿𝑆𝑠 – the set of lessons, which are held for student Ss. in the Eq. (10) the amount 𝐿𝑖𝑠 from the set 𝐿𝑙𝑒𝑐𝑡𝑢𝑟𝑒𝑠 ∩ 𝐿𝑇𝑡(𝑇𝑑𝑎𝑦)
𝑇𝑡 (𝑇𝑑𝑎𝑦 )
𝐿 – the set of lessons that take place in the time interval no more than 2.
Tt(Tday). imax – the serial number of the last lesson, imin – the 7) Some wishes of teachers:
serial number of the first lesson. Thus, in the Eq. (5) the To formulate this requirement, a teachers' wishes matrix
amount 𝐿𝑖𝑠 from the set 𝐿𝑆𝑠 ∩ 𝐿𝑇𝑡(𝑇𝑑𝑎𝑦) is equal to the Mdesires is considered.
expression (iminmax+1) if the number of “windows” is 0.
2) The lessons number of students per day should not ( )
 Pp , Tt : Pp  P; Tt  T ;
exceed the allowable number of lessons per day: (!L : (P = P ) (L = L ) (m = 1))
i i
*
p i
Tt
p, t (11)

(
 S s , Tt
day
): S  S ; Tt
day
( )
 T Tday ;  (! L : (P = P ) (L = L ) (m = 0))
i i
*
p i
Tt
p, t
s
(6)
 L( is
Tt Tday )
 N ld 𝑁𝑝 ,𝑁𝑡
𝑀𝑑𝑒𝑠𝑖𝑟𝑒𝑠 = (𝑚𝑝,𝑡 )𝑝=1,𝑡=1 , where mp,t=1, if there is a ban on
iL  L
Ss

the lesson for the p-th teacher during the t-th lesson, and mp,t=0,
𝐿𝑆𝑠 – the set of lessons, which are held for student Ss. if there is no ban.
𝐿𝑇𝑡 (𝑇𝑑𝑎𝑦 )
– the set of lessons that take place in the time interval 𝐿𝑇𝑡 – the set of lessons that take place in the time interval Tt.
𝑇𝑡 (𝑇𝑑𝑎𝑦 ). Thus, in the Eq. (6) the amount 𝐿𝑖𝑠 from the set 𝐿𝑆𝑠 ∩ The expression (11) means that there is a single cycle of
lessons where the participant of the educational process
𝐿𝑇𝑡(𝑇𝑑𝑎𝑦) no more than the allowable number of lessons for a conducts lessons in a given time interval, if there is no ban on
student per day Nld. lessons, or such a cycle does not exist and in this time interval
3) Correspondence of the type of audience to the type of lessons are not held. If the teacher's desire leads to non-
lesson: fulfillment of hard constraints, the desire will not be fulfilled,
or there will be another decision that does not violate hard
( Aa ) : Aa  A; constraints.
(
! Li : Likind = Ata ) (7) 8) Minimizing teachers’ transitions between buildings:

The expression (7) means that there is a single cycle of ( )

 Pp , Ttday : Pp  P; Ttday  T Tday ; ( )
lessons in which the type of audience is assigned to the
audience of the appropriate type.  Aib = N ab ;
i L
pp
L
(
Tt Tday )
(12)
(!L : ( = P ) (L = L ) (N  2))
4) The limits of hours per day do not exceed the
Tt
specified number of study hours: i Pi* p i ab

 (! L : (P = P )  (L = L ) (N  2 ))
( )
 S s , Ttday : S s  S ; Ttday  T Tday ; ( ) i i
*
p i
Tt
ab

(8)
 2(Lis )  N hd
T t T day
𝐿𝑇𝑡 – the set of lessons that take place in the time interval Tt.
iL  L
Ss
The expression (12) means that there is a single cycle of
lessons when the participant of the educational process
𝐿𝑆𝑠 – the set of lessons, which are held for student Ss. conducts lessons, where the number of educational buildings
𝐿𝑇𝑡(𝑇𝑑𝑎𝑦) – the set of lessons that take place in the time interval in which the audiences are located is not more than 2, or such
𝑇𝑡 (𝑇𝑑𝑎𝑦 ). Thus, in the Eq. (8) the amount 2𝐿𝑖𝑠 from the set a cycle does not exist.
9) Minimizing students’ transitions between buildings:
𝐿𝑆𝑠 ∩ 𝐿𝑇𝑡(𝑇𝑑𝑎𝑦) no more than a specified number of hours per
day for students Nhd.
5) Lectures preferably carried out in the first, second ( )
 S s , Ttday : S s  S ; Ttday  T Tday ; ( )
lessons:
 Aib = N ab ;
( )
( )
Tt Tday
i L SS
L (13)
 Tt pair : Tt pair  T pair ;
(!L : (S = S ) (L = L ) (N  2)) Tt
(( ) ( ))
*
(9) i i p i ab

 (! L : (S = S )  (L = L ) (N  2 ))
! Li : Likind = lecture  Tt pair  2 * Tt
i i p i ab
The expression (9) means that there is a single cycle of
lessons in which type of class is a lecture and the order of its 𝐿𝑇𝑡 – the set of lessons that take place in the time interval Tt.
placement in the time slot is not higher than the second lesson. The expression (13) means that there is a single cycle of
6) The number of lectures should not exceed two per day: lessons when the student has lessons, where the number of

380
buildings in which the audiences are located is not more than allowable number of lessons per day.
2, or such a cycle does not exist.
Cld
ld = 1 − max
→ max (16)
5. COURSES TIMETABLING CREATION PROCESS Cld
TARGET FUNCTION
The efficiency of schedule by constraint (7) – ϕat is
The schedule of courses should be as convenient as possible determined for each lesson and it takes value 𝜙𝑎𝑡 ∈ [0; 1]. The
for both students and participants in the learning process [7]. minimum possible number of discrepancies in the type of
𝑚𝑖𝑛
It is difficult to choose one criterion by which to assess the audiences is equal 𝐶𝑎𝑡 , and the maximum possible number
𝑚𝑎𝑥
quality and effectiveness of the schedule of courses. In section of discrepancies in the type of audiences is 𝐶𝑎𝑡 𝑙 . Then ϕat (17)
“Soft Constraints Description” a list of criteria that will be is the ratio of the difference between the maximum possible
used in the work to optimize the schedule of courses sessions number of discrepancies between the type of audience and the
𝑚𝑎𝑥
was proposed. type of lessons 𝐶𝑎𝑡 and the actual number of discrepancies
Let's assume that 𝜙𝑖 : 𝑋 → 𝑅1 , 𝑖 = 1, 𝐾 – functions, each of between the type of audience and the type of lessons 𝐶𝑎𝑡 to the
which optimizes one criterion for assessing the schedule of maximum possible number of discrepancies between the type
𝑚𝑎𝑥
training sessions, where K – this is the total number of criteria. of audience and the type of lessons 𝐶𝑎𝑡 . From this statement
Then the criterion of efficiency in the courses timetabling it follows that the greater the value ϕat, the higher the
creation task will be: efficiency of the created schedule of courses, because the
number of audiences that do not correspond to the type of
K lessons is less.
F =  i  i → max (14)
i =1 Cat
 at = 1 − max
→ max (17)
λi – a "weighting factor" that determines the significance of Cat
the i-th criterion. The "weighting factor" is determined by the
operator, before creating a schedule based on their own The efficiency of schedule by constraint (8) – ϕhd is
considerations about the importance of each efficiency criteria. determined for each student and it takes value 𝜙ℎ𝑑 ∈ [0; 1].
The difference between the current and the "ideal" schedule is The minimum possible number of exceedances of study hours
an integral assessment of the schedule by all criteria. per day for the allowable number of study hours per day is
𝑚𝑖𝑛
The efficiency of schedule by constraint (5) – ϕw is equal 𝐶ℎ𝑑 , and the maximum possible number of excesses is
𝑚𝑎𝑥𝑠 𝑑𝑎𝑦
determined for each student and it takes value 𝜙𝑤 ∈ [0; 1]. 𝐶ℎ𝑑 . Then ϕhd (18) is the ratio of the difference between
The minimum possible number of “windows” is 𝐶𝑤𝑚𝑖𝑛 , and the the maximum possible number of exceedances of study hours
maximum possible number of “windows” is per day for the allowable number of study hours per day 𝐶ℎ𝑑 𝑚𝑎𝑥
𝑖
𝑚𝑎𝑥𝑠 ((𝑇𝑝𝑎𝑖𝑟 ) ())
𝑖 and the actual number of exceedances of study hours per day
𝐶𝑤 𝑚𝑎𝑥
, where (𝑇𝑝𝑎𝑖𝑟 ) – the maximum for the allowable number of study hours per day Chd to the
𝑚𝑎𝑥
possible number of lessons per day. Then ϕw (15) is the ratio maximum possible number of exceedances of study hours per
𝑚𝑎𝑥
of the difference between the maximum possible number of day for the allowable number of study hours per day 𝐶ℎ𝑑 .
"windows" 𝐶𝑤𝑚𝑎𝑥 and the actual number of "windows" Cw to From this statement it follows that the greater the value ϕhd,
the maximum possible number of "windows" 𝐶𝑤𝑚𝑎𝑥 . From this the higher the efficiency of the created schedule of courses,
statement it follows that the greater the value KW, the higher because the number of exceedances of the allowable number
the efficiency of the created schedule of courses, because the of study hours per day is less.
number of "windows" is less.
Chd
Cw hd = 1 − max
→ max (18)
w = 1− → max (15) Chd
Cwmax
The efficiency of schedule by constraint (9) – ϕnl is
The efficiency of schedule by constraint (6) –ϕld is determined for each student and it takes value 𝜙𝑛𝑙 ∈ [0; 1].
determined for each student and it takes value 𝜙𝑙𝑑 ∈ [0; 1]. The minimum possible number of lectures held after the
The minimum possible number of exceedances of the number second lesson for each student is equal 𝐶𝑛𝑙 𝑚𝑖𝑛
, and the
𝑚𝑖𝑛
of lessons per allowable number per day is equal 𝐶𝑙𝑑 , and the maximum possible number of lectures held after the second
𝑁𝑙𝑑
𝑚𝑎𝑥𝑠 𝑑𝑎𝑦
𝑁 𝑠
𝑚𝑎𝑥 ∑𝑖=1 𝑖
𝑁𝑙𝑑 +1
lesson for each student is 𝐶𝑛𝑙 . Then ϕnl (19) is the ratio
maximum possible number of excesses is 𝐶𝑙𝑑 , of the difference between the maximum number of lectures
𝑙𝑑 𝑚𝑎𝑥
where 𝑁𝑖 – the number of classes for each student according given after the second lesson for each student 𝐶𝑛𝑙 and the
to the plan, N ld – the permissible number of classes for a actual number of lectures given after the second lesson for
student per day. Then ϕld (16) is the ratio of the difference each student Cnl to the maximum possible number of lectures
𝑚𝑎𝑥
between the maximum possible number of exceedances of the held after the second lesson for each student 𝐶𝑛𝑙 . From this
number of lessons for the allowable number 𝐶𝑙𝑑 𝑚𝑎𝑥
and the statement it follows that the greater the value ϕnl, the higher
actual number of exceedances of the number of lessons for the the efficiency of the created schedule of courses, because the
allowable number Cld to the maximum possible number of number of lectures after the second lesson is less.
exceedances of the number of lessons for the allowable
𝑚𝑎𝑥 Cnl
number 𝐶𝑙𝑑 . From this statement it follows that the greater  nl = 1 − → max
the value ϕld, the higher the efficiency of the created schedule max (19)
Cnl
of courses, because the less the number of exceedances of the

381
 The efficiency of schedule by constraint (10) – ϕal is The minimum possible number of transitions for students
determined for each student and it takes value 𝜙𝑎𝑙 ∈ [0; 1]. 𝑚𝑖𝑛
between buildings is 𝐶𝑠𝑑 , and the maximum number of
The minimum possible number of days where more than two transitions between buildings equal
𝑚𝑖𝑛
lectures for each student is 𝐶𝑎𝑙 , and the maximum possible 𝑚𝑎𝑥 ∑
∑
𝑆
(𝑖
𝑇𝑡 (𝑇𝑑𝑎𝑦 ) 𝑚𝑎𝑥 ())
number of days where more than two lectures for each student 𝐶𝑠𝑑 𝑖∈𝐿 𝑠 ∩𝐿 , where imax – this is the ordinal
𝑚𝑎𝑥𝑠 𝑑𝑎𝑦
is 𝐶𝑎𝑙 . Then ϕal (20) is the ratio of the difference of the number of the last lesson of the student on a particular day.
maximum number of days where the lectures are more than Then ϕsd (23) is the ratio of the difference between the
two for each student 𝐶𝑎𝑙 𝑚𝑎𝑥
and the actual number of days maximum number of transitions of students between buildings
𝑚𝑎𝑥
where there are more than two lectures for each student Cal to 𝐶𝑠𝑑 and the actual number of students transitions between
the maximum possible number of days, where there are more buildings Csd to the maximum possible number of transitions
𝑚𝑎𝑥
than two lectures for each student 𝐶𝑎𝑙 𝑚𝑎𝑥
. From this statement of students between buildings 𝐶𝑠𝑑 . From this statement it
it follows that the greater the value ϕal, the higher the follows that the greater the value ϕsd, the higher the efficiency
efficiency of the created schedule of courses, because the of the created schedule of courses, because the number of
number of days where the number of lectures is more than two transitions of students between buildings is less.
is less.
Csd
 sd = 1 − max
→ max (23)
Cal Csd
 al = 1 − max
→ max (20)
Cal
This approach is not the only one. The number of constraints
The efficiency of schedule by constraint (11) – ϕd is may increase, but this will increase the complexity and
determined for each teacher and it takes value 𝜙𝑑 ∈ [0; 1]. flexibility of the system. There may be problems with the
The minimum possible number of non-fulfillment of teachers' solution of large systems, namely: lack of computer memory,
loss of accuracy, the way to improve results will be very
wishes is equal 𝐶𝑑𝑚𝑖𝑛 , and the maximum possible number of
𝑁𝑝 difficult and long.
𝑚𝑎𝑥 ∑
𝑖=1 𝑑𝑒𝑠𝑖𝑟𝑒𝑠
𝑁𝑖
non-fulfillment of teachers' wishes is 𝐶𝑑 , where
𝑖
𝑁𝑑𝑒𝑠𝑖𝑟𝑒𝑠 – the number of wishes of the i-th teacher. Then ϕd
(21) is the ratio of the difference between the maximum 6. EFFICIENCY AND FINDING A COMPROMISE
number of non-fulfillment of teachers' wishes 𝐶𝑑𝑚𝑎𝑥 and the BETWEEN THE CRITERIA
actual number of non-fulfillment of teachers' wishes Cd to the
maximum possible number of non-fulfillment of teachers' Analysis of the criteria for optimality of schedule does not
allow to identify the most important of them, assuming that
wishes 𝐶𝑑𝑚𝑎𝑥 . From this statement it follows that the greater
others are insignificant [12]. In cases where the efficiency
the value ϕd, the higher the efficiency of the created schedule
indicator cannot be represented as one indicator and one
of courses, because the number of unfulfilled wishes of
optimality criterion, the problem is called multicriteria. In the
teachers is less.
case of a multicriteria problem, the optimization problem is
considered as a problem of finding a compromise between the
Cd
d = 1 − → max (21) criteria [13]. These criteria may reflect assessments of the
Cdmax different qualities of the object or process about which the
decision is made [14].
The efficiency of schedule by constraint (12) – ϕpd is One type of compromise is the Pareto compromise. In this
determined for each teacher and it takes value 𝜙𝑝𝑑 ∈ [0; 1]. case, we get not one solution, but the solution area of the
The minimum possible number of transitions for teachers optimization problem. Defining the area of compromises Qx is
𝑚𝑖𝑛
between buildings is 𝐶𝑝𝑑 , and the maximum number of to select from a set of acceptable solutions Q a subset having
the property that each solution 𝑞 ∈ 𝑄𝑥 , cannot be improved
transitions between buildings equal
∑ without deteriorating at least one of the criteria. [15].
𝑚𝑎𝑥 ∑ (𝑖
𝑇𝑡 (𝑇𝑑𝑎𝑦 ) 𝑚𝑎𝑥 ())
𝑃𝑝 On the set of variants of change of the schedule, at its
𝐶𝑝𝑑 𝑖∈𝐿 ∩𝐿 , where imax – this is the ordinal
optimization, the set of Pareto-optimal variants is defined, i.e.
number of the last lesson of the teacher on a particular day. options where the integrated estimate of the schedule is close
Then ϕpd (22) is the ratio of the difference between the to zero, i.e. the estimate of the "ideal" and the actual schedule
maximum number of transitions of teachers between buildings are equal, or almost equal, in other words, a schedule that does
𝑚𝑎𝑥
𝐶𝑝𝑑 and the actual number of teachers transitions between not deteriorate compared to the original schedule. Such
buildings Cpd to the maximum possible number of transitions decisions include those where deterioration by one criterion
𝑚𝑎𝑥
of teachers between buildings 𝐶𝑝𝑑 . From this statement it does not result in a weakening by any other criterion.
follows that the greater the value ϕpd, the higher the efficiency Ideally, the task of forming a schedule of courses can be
of the created schedule of courses, because the number of considered as a task of finding such 𝑥 ∈ 𝑄 , at which the
transitions of teachers between buildings is less. maximum of all private criteria is reached, where x – allowable
schedule, Q – the set of all possible allowable schedules [16]:
C pd
 pd = 1 − → max (22)
C max  i ( x) → max
pd  (24)
x  Q
The efficiency of schedule by constraint (13) – ϕsd is
determined for each student and it takes value 𝜙𝑠𝑑 ∈ [0; 1]. In practice, a solution in which the maximum of all private

382
criteria is achieved simultaneously exists as an exception. This In general, to maximize indicators F1(x) and F2(x) is
is often due to various contradictions between the subjects of contradictory. For example, changes in some schedule x,
the schedule. Therefore, the solution of such problems should increasing the indicator F1(x), can lead to serious disturbances
be sought in the set of Pareto-optimal solutions 𝑄𝑥 ⊂ 𝑄. Then in the schedule of a subject that will lead to increased indicator
choose the resulting schedule from the set Qx could be F2(x), this is especially evident in problems of medium and
performed by an expert, however, the power of the set Qx for large size with a large number of requirements for the schedule.
problem (24) can be large, which makes its analysis by an Figure 2 shows a diagram for calculating the effectiveness
expert difficult and tedious. Therefore, it is better to reduce the of the schedule and checking for compliance with hard
number of criteria for evaluating schedules. The problem of constraints.
determining the set Qx takes the form: From the output of the "Schedule Creator" block, a ready-
made schedule is sent to the input of the "Hard Constraints
 Fk (x ) → max, k = 1,2 Check (HCC)" block, which is checked for compliance with
 K hard constraints. If at the outputs of the "Hard Constraints
 F (x ) =    (x )
 1  i i
Check (HCC)" block all values are equal to 1, then all
 i =1 (25) constraints are met and the quality of the schedule can be
 F (x ) = min (   (x )) calculated. The input of the "Schedule Efficiency" block
 2 i =1...K
i i
receives the schedule and the response that all the prerequisites
x  Q
 are met. The block "Schedule Efficiency" calculates private
estimates of the effectiveness of the schedule and the general
where, Fk(x), k=1, 2 – private criteria indicators of the problem, estimate of the schedule. The results of several experiments
F1(x) – general assessment of the schedule of courses x; F2(x) are listed in Table 1.
– minimum (worst) individual assessment of the schedule x.

Figure 2. Simulink scheme for calculating schedule efficiency indicators

Table 1. The results of experiments

№ F(x) φw φld φat φhd φnl φal φd φpd φsd

1. 7.74 0.85 0.82 0.91 0.86 0.88 0.9 0.63 0.99 0.9
2. 8.04 0.96 0.82 0.99 0.95 0.85 0.86 0.7 0.95 0.96
3. 8.08 0.93 0.97 0.94 0.83 0.9 0.79 0.94 0.86 0.92
4. 8.41 0.96 0.81 0.97 0.95 0.99 0.92 0.99 0.93 0.89
5. 8.54 0.96 0.87 0.95 0.99 0.99 0.96 1 0.94 0.88
Notes: 1. The number of checked schedules is 5. 2. Minimum individual grades are in bold and italicized.

383
 Analyzing the results obtained for the first experiment, we arrangement with applied genetic algorithm. 2009 Fourth
are looking for a private estimate with the lowest value. By International Conference on Computer Sciences and
optimizing the schedule by increasing the value of the Convergence Information Technology, pp. 1266-1270.
minimum estimate, we observe the value of the total estimate https://doi.org/10.1109/iccit.2009.202
of the schedule. We see that in the third variant, by increasing [7] Lytvynenko, O., Kralina, H., Stopushkina, O. (2004).
the previous estimate of the schedule, another partial estimate Matematychna model zadachi skladannia rozkladu
decreased. By increasing this estimate, another particular navchalnykh zaniat. Proceedings of National Aviation
estimate worsens. By the Pareto principle, we are looking for University. 9(1): 180-186.
a compromise between the schedule estimate and the worst https://doi.org/10.18372/2306-1472.19.992
private estimate. [8] Hasuhadzhiev, A.S. (2017). Formirovanie sistemy
pokazatelej dlja avtomatizacii uchebnogo raspisanija
tipovogo vuza. Vestnik AGTU. Serija: Upravlenie,
7. CONCLUSIONS vychislitel'naja tehnika i informatika, 117-127.
https://doi.org/10.24143/2072-9502-2017-3-117-127
The paper describes a mathematical model of courses [9] Khabipov, R. (2018). Avtomatizatsiya postroyeniya
timetable creation process. During this, in the first stage, all raspisaniya zanyatiy v vuze: matematicheskaya model i
process variables were identified, namely lessons, disciplines, metody realizatsii. Elektronnyee biblioteki, 21(5): 461-
teachers, students, audiences and timeslots. The next step, 470. https://elbib.ru/article/view/480/559, accessed on
using the described variables, was a description of hard and Nov. 11, 2019
soft constraints. The importance of describing them is that [10] Kabalnov, Yu., Shekhtman, L., Nizamova, G.
non-compliance with hard constraints will result in the student Zemchenkova, N. (2006). Kompozitsionnyy
not fully fulfilling the curriculum, and soft constraints are geneticheskiy algoritm sostavleniya raspisaniya
needed to improve the efficiency and quality of the courses uchebnykh zanyatiy. Vestnik UGATU. 7(2): 99-107.
schedule. Therefore, a number of performance criteria were http://journal.ugatu.ac.ru/index.php/Vestnik/article/view
described, according to which the evaluation of the obtained /1354, accessed on Nov. 11, 2019.
schedule will be performed. But due to the fact that in practice [11] Verevkin, V., Ismagilova, O., Atavin, T. (2009).
the schedule in which the maximums of all criteria are reached Avtomatizirovannoye sostavlenyee raspisaniya
exists as an exception, due to the contradictions of the subjects. uchebnykh zanyatiy vuza s uchetom trudnosti distsiplin i
Therefore, the number of criteria was reduced to two. The utomyeyaemosti studentov. Doklady TUSUR. 1(19):
solution of such problems should be sought in the set of 221-225.
Pareto-optimal solutions. https://cyberleninka.ru/article/n/avtomatizirovannoe-
The obtained results will be used in the description of the sostavlenie-raspisaniya-uchebnyh-zanyatiy-vuza-s-
algorithm and the development of an automated control uchetom-trudnosti-distsiplin-i-utomlyaemosti-studentov,
system for the courses timetable creation process. accessed on Jan. 15, 2020.
[12] Klevanskiy, N.N., Puzanov A.A., Rubtsov, O.G. (2006).
Modeli i metody mnogokriterialnoy optimizatsii
REFERENCES raspisaniy vuza. Vestnik Saratovskogo
gosudarstvennogo tekhnicheskogo universiteta, 17(2):
[1] Sakaliuk, O., Trishyn, F. (2019). Analysis of process 77-83. http://oldlib.sstu.ru/open/vestniki/2006/04_17_-
creation of the courses timetabling. Automation of 2006.pdf, accessed on Feb. 9, 2021.
Technological and Business Processes, 11(2): 30-35. [13] Khobin, V.A. (2008). Sistemy garantiruyushchego
https://doi.org/10.15673/atbp.v11i2.1370 upravleniya tekhnologicheskimi agregatami: Osnovy
[2] Petrovic, S., Burke, E. (2004). University timetabling. teorii, praktika primeneniya. Odessa: Odesskaya
Handbook of Scheduling: Algorithms, Models, and natsionalnaya akademiya pishchevykh tekhnologiy, 1-
Performance Analysis, Chapter 45: 1-34. 308.
[3] Adewumi, A., Sawyerr, B., Montaz Ali, M. (2009). A [14] Chybisov, Y.V., Shulha, Y.S. (2014). Zastosuvannia
heuristic solution to the university timetabling problem. metodiv bahatokryterialnoi optymizatsii dlia vyrishennia
Engineering Computations, 26(8): 972-984. zadachi rozpodilu vahoniv po vantazhnym frontam.
https://doi.org/10.1108/02644400910996853 Transportni Systemy Ta Tekhnolohii Perevezen, 7(1):
[4] Lih, B.C., Nah, S.S., Bolhassan, N.A. (2015). A study on 65-72. https://doi.org/10.15802/tstt2014/35994
heuristic timetabling method for faculty course timetable [15] Kostenko, A.N. (2000). Opredeleniye oblasti
problem. 2015 9th International Conference on IT in kompromissov pri reshenii zadach optimizatsii v
Asia (CITA), pp. 1-3. mnogokriterialnoy postanovke. Sistemi Obrobki
https://doi.org/10.1109/cita.2015.7349832 Іnformatsії 2(8): 52-56
[5] Hooshmand, S., Behshameh, M., Hamidi, O. (2013). A http://www.hups.mil.gov.ua/periodic-app/article/13004,
tabu search algorithm with efficient diversification accessed on Feb. 15, 2021.
strategy for high school timetabling problem. [16] Bezginov, A., Tregubov, S. (2011). Mnogokriterialnyy
International Journal of Computer Science and podkhod k otsenke raspisaniya zanyatiy na osnove
Information Technology, 5(4): 21-34. nechetkoy logiki. Problemy upravleniya, 2: 52-59.
https://doi.org/10.5121/ijcsit.2013.5402 https://cyberleninka.ru/article/n/mnogokriterialnyy-
[6] Khonggamnerd, P., Innet, S. (2009). On improvement of podhod-k-otsenke-raspisaniya-zanyatiy-na-osnove-
effectiveness in automatic university timetabling nechyotkoy-logiki, accessed on Feb. 9, 2021.

384
NOMENCLATURE S set of students
T set of timeslots
A set of audiences x schedule of courses
D set of disciplines
F process target function Greek symbols
K total number of criteria
L set of lessons λ weighting factor that determines the degree of
M teachers' wishes matrix importance of the criterion
N number of the elements in the set φ assessment that determines non-compliance with
P set of participants in the educational process the desired requirement or soft constraints
Q set of all possible allowable schedules

385