Simulation of University Education Process

Jiří Jelínek

Institute of Applied Informatics, University of South Bohemia, Branišovská 31, České Budějovice, Czech Republic

Keywords: Agents, Simulations, University Education, Education Process Quality.

Abstract: The article deals with one possible usage of agent based simulation. This method is advantageous for tasks

that can be modeled through a set of basic building elements - agents and their interaction. This contribution

shows an application of this approach for the simulation of university education. Conceptual model of

university education based on AB modeling of the students was developed for this purpose. The aim was to

develop a tool which could help to improve the quality of higher education processes through better

knowledge about them obtained from their simulations.

1 INTRODUCTION

Solving problems using simulation techniques has

become a very popular method particularly in recent

years. Three main approaches to simulation are

available to us. The first one which is based on

system dynamics (SD) has been used since the

1960s, and it was first introduced by Jay Forrester

(Forrester, 1961). This is the macro view of problem

to be solved.

The second one is based on the occurrence of

specific events (DE - Discrete Event). The modeled

system is described using activities that affect states

of the system.

Building the model consisting of similar,

relatively simple objects is the third option. Their

behavior and mutual interactions create global

characteristics of the system, which we call agent

based (AB) approach.

This paper will show an application of AB

simulation in the area of higher education with a

conceptual model and its implementation presented.

The aim of the work was to create a model of a

university study process based on the credit system,

which would be usable for simulations related to the

preparation of teaching methodology and subsequent

improvement of its quality.

Section 2 of this article presents related work,

and the solution to this problem is described in

section 3. Section 4 is the core of the paper and here

the conceptual model of university education is

presented. Chapter 5 is focused on the

implementation of the model and conclusions are

presented in section 6.

2 CURRENT STATE

Creating AB simulations is a hot topic, mainly

because of availability of tools applicable to this

area, with some of them being e.g. NetLogo

(Wilensky, 1999), REPAST (Argonne National

Laboratory, 2012) and Swarm (Swarm Development

Group, 2012). These systems are freely available on

the Internet. Commercially available, AnyLogic

(AnyLogic Company, 1992-2012) allows to

combine several above mentioned approaches to

create heterogeneous models. The Izumi’s paper

(Izumi, et al., 2007) is suitable introduction to the

AB approach. Information about AB approach can

also be found in Siebers (Siebers & Aickelin, 2008).

The use of AB approach for university study

modeling has not been specifically published in

scientific literature and therefore we can conclude

that it is a little explored area.

Problems of coordination and behavior of agents

in the socio-technical systems are discussed by

Eccles et.al. (Eccles & Groth, 2006). The Tang’s

article (Tang, et al., 2006) also touches on this

theme, but its content is primarily focused on

modeling human capital hidden in education.

Additional materials are focused on exploring

interactions in social networks, such as (Komis, et

al., 2002). Modeling of certain student properties

(namely loyalty) is described in an article (Helgesen

& Nesset, 2009). Exploration of collective behavior

403

Jelínek J..

Simulation of University Education Process.

DOI: 10.5220/0004258904030406

In Proceedings of the 5th International Conference on Agents and Artiﬁcial Intelligence (ICAART-2013), pages 403-406

ISBN: 978-989-8565-38-9

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

is explored, for instance, in (Xiang-Min & Ming-

Yong, 2010). An interesting approach is presented in

(Ayala, 2009), where the student model is based on

the definition of appropriate ontology. Modeling a

particularly mental capacity of a student is discussed

in (Dimitrova, 2003). Ammar (Ammar, et al., 2010)

comments the incorporating emotional stimuli and

on inputs into the educational system.

3 THE UNIVERSITY STUDY

PROCESS

For simulation the study process was modeled in 4

steps.

The admission process, being an important

information source for setting up the model is first.

The second step is the student’s enrollment in

courses each semester. The key is the process of

selecting a course and fulfilling the necessary

prerequisites as stated by the requirements of the

particular major.

The course study process is not important for the

modeling purposes. Only the third step is essential -

study results and how they are determined.

After the study period, certain administrative

tasks have to be completed and these have a direct

bearing on the student’s progress. Afterwards the

student can enroll in other courses in the next study

period or terminate (successfully or unsuccessfully)

his/her study. This is the fourth step.

The proposed study process is based on the credit

system and assumes the existence of three groups of

courses: required, compulsory elective and optional.

The student must obtain a specified number of

credits from each of these groups. Minimum number

of credits that the student must acquire in the study

period is also a certain motivation for him/her.

4 THE CONCEPTUAL MODEL

The task described in chapter 3 is an ideal problem

for the AB approach because many very autonomous

and (in principle) similar objects participate in it.

An obvious candidate for modeling with the AB

approach is the student himself. But the question is

whether such an autonomous unit is also the course

(including its teacher) itself. The introduced model

does not assume that, with the courses being treated

as passive objects with given properties and

methods.

4.1 Model Environment

Environment that will define the basic simulation

parameters is given by rules and regulations of the

school. Its data structures must therefore store a

rating scale, setting of the assessment grades, as well

as recommended schedules (plans) of study. The

basic time unit is one semester (one study period).

A list of subject matters T used as coordinates in

descriptive vectors for the students and courses is

essential for the model. Number of subject matters is

not fixed and depends on the desired degree of detail

and the ability to identify coordinate values. For

simplicity and testing, the whole course can be

replaced by one subject matter.

A very important parameter for assessing the

quality of the learning process is the ideal graduate

profile vector Q.

The main method associated with the model

environment is, without doubt, model time control

and the related management of actions in the process

of simulation. The DE approach is used.

4.2 Simulation of the Student

Each student is modeled by his/her attributes and

behavior. The basic idea is to represent the student

in the model with four vectors with dimensions

defined by T which describe a student person each

from a different view (Eq.1). Coordinates of these

descriptive vectors are normalized in the range < 0,

1 > (1 means highest or best).

The vector of knowledge K, which stores the

actual level of student knowledge of the subject

matter, is first. The second vector deals with

personal preconditions A (study skills). It is assumed

that the student has different dispositions to study

various subjects which do not change in time. The

third vector is the motivation vector M. Again,

different values of motivation for different subjects

are assumed. The last is the goals vector G. This

reflects the student's preference for knowledge

he/she wants to acquire in school, in other words

his/her professional focus.

Basic characteristics of the student can also be

written in a matrix (Eq.1).





















.........

(1)

ICAART2013-InternationalConferenceonAgentsandArtificialIntelligence

404

The initial values of the vectors K and A should

be taken from the results of the admission tests. The

coordinates of the vector K will increase

monotonously. On the other hand values of vector M

may change in both directions. The values of vector

G may vary during the study according to the

changing preferences of the student affected by both

internal and external factors.

In addition to these basic data structures, each

student is assigned a list of courses S, which he/she

is attending during the current semester and an

archive of finished courses.

In addition to the above attributes, it is necessary to

equip the student with procedural knowledge and

skills, i.e. definition of his/her behavior (methods).

One of the basic methods is undoubtedly the

selection of student courses for the entire semester.

The selection depends on the values of vectors G, A,

K and M, on course credits and completed courses,

which may constitute the prerequisites or must be

repeated. Other aspects that influence the selection

are: the requirement factor of the course, number of

credits obtained from groups of courses and

available space in the course.

The second group of methods is focused on the

evaluation of the course and on the changes of the

knowledge vector K and motivation vector M. These

methods are usually utilized after the completion of

the entire semester. Because of the interaction with

the course description, more detail is given below.

4.3 The Course Model

The course is modeled by vectors with the same

dimensions as the vectors describing the student.

Input vector I, which expresses the level of required

skill for entering and completing the course, is the

first vector. The second vector O, the output vector,

defines the maximum level of knowledge the student

can achieve by studying the course. The course

description could be written as an Eq. 2.













.........

(2)

The course model also contains additional data

structures (e.g. the list of preconditions, the

definition of duty or number of credits). An

important limiting factor is the capacity of the

course and its related actual occupancy.

The amount of knowledge gained by the student

is probably the most important method associated

with the course.























tst

tsttsttttSs

kokoamg

,max

(3)

As shown in Eq. 3, the change of coordinates of

vector K is calculated as the maximum from

expansion of knowledge across all courses studied

during the entire semester S using the difference

between the maximum possible output value of the

course o

and the original value k

and the

coordinates g

, m

and a

of corresponding vectors

and it reflects the influence of study goals,

motivation and study skills relevant to the subject

matter. This procedure takes into account the mutual

influence of courses studied in the same period.

A related method to the above mentioned one is

evaluation of the student. The evaluation c

of the

student in the course s is the measure of success of

the student in the course according to Eq. 4 and the

specific grade is derived from the study regulations

(e.g. c

= 0.6 to c

= 0.75 means grade C).



stt

sttst

stt

stst

oki



























(4)

The i

, o

and k

are the coordinates of the

corresponding vectors.

For the proper functionality of the model it is

necessary to define the change of motivation vector

M after the semester ends. The model assumes that

the student is positively motivated by extreme study

results, both positive and negative. In contrast,

average results have a negative motivational effect.

Motivation Δm

increases if the resultant evaluation

from the course s in selected subject matter t is far

enough from the average results of the student,

otherwise it declines. The final change Δm

of the

motivation vector coordinate is the maximum of all

the changes Δm

across all courses s studied in the

entire semester S.

5 EXPERIMENTS AND FUTURE

WORK

The above study model was implemented in

AnyLogic and was verified by testing data regarding

the study of approximately 200 students in 60

courses during 9 semesters. Courses included 11

subject matters. The model provides both online

SimulationofUniversityEducationProcess

405

summary information, as well as storage of

individual student data for offline use.

The model computes the quality q and also q

all subject matters (Eq. 5) for every student. Vector

Q is the graduate profile vector, and K is the

knowledge vector. The quality values are important

for the evaluation of effectiveness and success of the

study process and the quality of the graduates.

q  ,

(5)

Experiments which were performed were designed

to validate the model using real process of students’

studies. Due to the structure of the admission

procedure it was not possible to use data from it for

setting the student parameters, and initial values of

vectors K, A, G, and M were thus generated from

estimated intervals with Gaussian probability

distributions. The values of course vectors I and O

were obtained from discussions with students. Initial

results confirm the hypothesis, that especially

preconditions together with capacity limits affect the

structure of the student’s study plans.

The main goal of future work is to specify

parameters of the model from data in the faculty

information system and from data contained in the

admissions results.

Our first goal is to determine the parameters for the

probability distributions used for setting of the

student simulation vectors K, A, M a G. The model

can then be used for exploring global study

processes.

In the second step we would like to suggest

mechanisms enabling the use of the proposed system

for modeling real learning process of the students by

utilizing regularly updated data. This would allow

prediction of their study results and limit potentially

problematic situations.

6 CONCLUSIONS

This paper describes the agent-based model for the

simulation of the process of university studies. Its

aim is to contribute to the effectiveness and setting

of study plans and processes. The first experiments

were conducted with the model to verify its

applicability for solving practical problems. Further

development will focus mainly on validating and

fine-tuning the model for accurate simulation of real

world situations and verifying its predictive

capability.

REFERENCES

Ammar, M. B., Neji, M., Alimi, A. M. & Gouardères, G.,

2010. The Affective Tutoring System. Expert Systems

with Applications: An International Journal archive,

37(4), pp. 3013-3023.

AnyLogic Company, 1992-2012. AnyLogic. [Online]

Available at: http://www.anylogic.com

[Accessed 2. 11. 2012].

Argonne National Laboratory, 2012. Repast. [Online]

Available at: http://repast.sourceforge.net/

[Accessed 15. 4. 2012].

Ayala, A. P., 2009. Student Modelling Based on

Ontologies. ACIIDS '09 Proceedings of the 2009 First

Asian Conference on Intelligent Information and

Database Systems, pp. 392-397.

Dimitrova, V., 2003. STyLE-OLM: Interactive Open

Learner Modelling. International Journal of Artificial

Intelligence in Education, 13(1), pp. 35-78.

Eccles, D. W. & Groth, P. T., 2006. Agent coordination

and communication in sociotechnological systems.

Interacting with Computers, 18(6), pp. 1170-1185.

Forrester, J. W., 1961. Industrial dynamics. Cambridge,

MA: MIT Press.

Helgesen, O. & Nesset, E., 2009. Modelling and

Managing Student Loyalty: A Study of a Norwegian

University College. Scandinavian Journal of

Educational Research, 53(4), p. 327 – 345.

Izumi, K., Matsui, H. & Matsuo, Y., 2007. Socially

embedded multi agent based simulation of financial

market. Proceedings of the 6th international Joint

Conference on Autonomous Agents and Multiagent

Systems (Honolulu, Hawaii, May 14 - 18, 2007)..

Komis, V., Avouris, N. & Fidas, C., 2002. Computer-

Supported Collaborative Concept Mapping: Study of

Synchronous Peer Interaction. Education and

Information Technologies, 7(2), pp. 169-188.

Siebers, P. O. & Aickelin, U., 2008. Introduction to Multi-

Agent Simulation. In: F. Adam & P. Humphreys, eds.

Encyclopedia of Decision Making and Decision

Support Technologies. Pennsylvania: Idea Group

Publishing, pp. 554-564.

Swarm Development Group, 2012. SWARM. [Online]

Available at: http://www.swarm.org

[Accessed 10. 5. 2012].

Tang, Y., Parsons, S. & Sklar, E., 2006. Agent-based

modeling of human education data. Proceedings of the

fifth international joint conference on Autonomous

agents and multiagent systems, pp. 129-131.

Wilensky, U., 1999. NetLogo. [Online]

Available at: http://ccl.northwestern.edu/netlogo/

[Accessed 22. 8. 2012].

Xiang-Min, G. & Ming-Yong, P., 2010. Modeling and

simulating dynamic evolvement of collective learning

behaviors by Voronoi diagram. Proceedings

LSMS/ICSEE'10, pp. 548-554.

ICAART2013-InternationalConferenceonAgentsandArtificialIntelligence

406