A NEW FRAMEWORK FOR THE CONTROL OF LMS IN ITS

Veronica Amela, José Luís Díez and Marina Valles

Departamento de Ingeniería de Sistemas y Automática, Universidad Politécnica, 46022 Valencia, Spain

Keywords: Intelligent tutoring systems, Learning management systems, Engineering education, Control education.

Abstract: Intelligent Tutoring Systems (ITS) offer an automatically personalized environment guiding the student and

allows him to put his knowledge and skills in a more effective way than with traditional lessons. Besides,

this learning methodology gives study location and time freedom to the student thanks to Internet facilities.

Moreover the Learning Management System (LMS) or plat-form which holds the ITS, gathers the course

materials and student information making them available and reusable for other control courses. In this

paper, we propose a new approach to LMS in ITS, applying data mining techniques.

1 INTRODUCTION

European syllabus reform for the new university

degrees implies a drastic change in the studies

organization. This reform considers the personal

time study as teaching hours.

One solution to put this reform into practice is,

for example, thanks to internet-based education. But

we must tutor the student in order to enhance his

learning process. Human tutoring is extremely

laborious and expensive; however ITS are computer

systems for custom-made learning, that don’t need

human tutor intervention.

This paper shows a new framework for ITS

applied to Control Systems and Automatic

Engineering studies.

2 MODEL

The ITS architecture (Fig. 1a) is organized in three

submodels based on three types of knowledge (Ong

and Ramachandran, 2003): Student model, Tutor

model and Domain model.

Making a comparison with control theory, the

process to be controlled in an ITS is the student

learning process.

The student shows a learning style and some

previous knowledge that are going to be dynamically

modified by the student interaction with the course.

The sensor system consists of information

extraction and storage in a database. Subsequently

this information is analyzed by data mining

techniques.

Data mining feedbacks the ITS controller. It

adapts the course's LMS models. The teacher

supervises the process and he defines the course

contents according to the course curriculum.

Apart from this first control loop that has just

been described (Figure 1a), it exists another slower

external loop in the model proposed (Figure 1b). It

verifies how well the ITS works and it controls how

fine it fits the student needs by comparing the

students results with the initial course objectives.

Figure 1a: ITS components.

Figure 1b: ITS model architecture.

The LMS model suggested (Figure 2) is made up

at the same time of student model, tutor model and

domain model. It also considers the relationships or

287

Amela V., Díez J. and Valles M..

A NEW FRAMEWORK FOR THE CONTROL OF LMS IN ITS.

DOI: 10.5220/0003062202870289

In Proceedings of the International Conference on Knowledge Management and Information Sharing (KMIS-2010), pages 287-289

ISBN: 978-989-8425-30-0

 2010 SCITEPRESS (Science and Technology Publications, Lda.)

Figure 2: Integration of the three models (Student, Tutor and Domain) inside the LMS.

information flows among them. The arrows

symbolize these relationships: red (system’s

initialization), black (direct feedback or information

update) and dashed line (initial student data taking

from surveys).

- Student Model: The first block models the

static features that don’t change in the course of time

(cognitive characteristics and learning preferences).

The other one gathers the dynamic features (student

knowledge and motivation state). This block is in

continuous updating because it depends on students

interaction with the LMS course.

- Domain Model: One course is organized in

didactic units. At the same time, these units are

formed by learning objects defined with metadata

(LOM: activities, workshops, tasks, multimedia

resources, laboratory experiences, etc.) LOM choice

and sequence for a learning route is made in the

Tutor model based on the course curriculum.

Student’s interaction with the course activities or

resources produces a series of reports or records (log

files). The system stores the log files together with

the student marks (they are got by expert and student

response comparison). By analyzing this

information the system sets the student progress.

The progress can be split in achievement or success

that will affect the student motivation and

knowledge level acquired to date.

- Tutor Model: At the beginning, it chooses from

the student model data (learning style) the

appropriate pedagogical methodology. The learning

style also establishes which kind of multimedia

materials the student prefers, which together with

the contents that should be learnt in the course

(knowledge level to be achieved) enable the LOM

choice that better fits the student’s needs. These two

blocks are suitable for planning the learning route.

A temporal LOM sequence for a lesson or didactical

unit is what we know as learning route.

- ITS Control Loop: The ITS model is formed by

the LMS model plus a control loop with the

objectives and results feedback (Figure 2).

3 DATA MINING

Now, we are going to describe the different sources

or tools from where we get the student data. This

data will need an analysis by means of data mining

techniques to extract useful information.

The Index of Learning Styles (ILS) questionnaire

is an on-line instrument used to assess preferences

on four dimensions of a learning style model (Felder

and Silverman, 1988). The ITS clusters the

questionnaire (data mining 1 in Fig. 1B) in order to

establish how strong the dimension is shown in the

student.

Knowledge level is based on Bloom taxonomy

(Blom, 1956) mixed with the collaborative

competence (Baldiris et al., 2008) that classifies the

students into different levels. There are several

studies about student’s academic aims and their

influence on motivation (González et al., 1996).

Deciding factors to deduce motivational

guidance are: social recognition that determines

extrinsic motivation, educational that determines

intrinsic motivation and Interpersonal that affects

student collaborative competence.

The Teenagers Goals Questionnaire (CMA in

Spanish) evaluates these 3 factors (Martín-Albo et

al., 2007). In the extrinsic case, we should determine

as well, the student confidence in his studies (self-

KMIS 2010 - International Conference on Knowledge Management and Information Sharing

288

esteem). This factor has an influence on the higher

difficulty level that the student is able to reach

without fail in his motivation.

The Rosenberg self-esteem scale (RSES)

evaluates the self-esteem in the university context

(Wolpers et al., 2007). Web-based educational

systems can record student’s system access in log

files.

Theses files provide a basic Internet student

tracking. An open code specification based in

attentionXML, is called Contextualized Metadata

Attention, CAM (Perkins, 1995). It captures user’s

observations (browser information use, Web sites,

news feeds, blogs, etc.)

The system clusters Student’s dataset (data

mining 4) to extract the different types of students

according to their knowledge level and learning

preferences. In this step, the system also looks for

anomalous values that correspond with

learninghandicaps or misunderstandings.

There are different pedagogical methodologies

suggested by several authors. Each of them has a

series of characteristics as regards to the type of

resources, student’s level of participation, student’s

learning involvement, etc. This choice entails a

specific learning route that can be associated (data

mining 5 in Fig. 1b and 2) with a unique student’s

profile.

The Tutor model organizes the units following

an established plan designed by the teacher who

bases his opinion on curriculum’s guidelines.

Materials chosen for each learning route will change

according to their success (voting), student difficulty

level and student pedagogical methodology. LMS

control system assessment and satisfaction is carried

out making a comparison between the student’s

average results during the LMS course and those of

the previous year, without the LMS platform.

On the other hand, a reduction in the subject

desertion rate also denotes the smooth running of the

system.

Finally, if the LMS course is implemented in all

subjects of the Engineering Grade, we can use the

University’s efficiency rate. It is defined as the

proportion between the total number of credits of the

grade syllabus and the total number of credits

enrolled by the student.

4 CASE STUDY

To date, it has been implemented the first part of the

Model proposed. It covers the static student model

analysis, applying data mining techniques, which

provide the type of students.

The initial surveys are facilitated to the students

thanks to open source application called

LimeSurvey. Student’s data is clustered by Fuzzy k-

prototypes algorithm (Ng and Wong, 2002) for

numerical and symbolic data developed in Matlab.

The result of the data mining process is the types

of students’ clusters. Once the clusters are made, by

means of a fuzzy rules algorithm (Nguyen and

Walker, 2000), we can associate each of them to the

pedagogical methodologies and resources’ difficulty

level. With these two factors, we can form the

adaptive and personal unit content sequence called

learning route for each student.

5 CONCLUSIONS

This work has focused on developing a new model

for ITS. The future work will finish the ITS model

implementation in a platform or LMS.

REFERENCES

Ong J. and Ramachandran S. 2003. Intelligent Tutoring

Systems: Using AI to Improve Training Performance

and ROI, Slotter Henke Associates

Felder, R. and Silverman, L. 1988. Learning and Teaching

Styles in Engineering Education. Engineering

Education, 78, Vol. 7, 674-881

Blom, B. S. 1956. Taxonomy of Educational Objectives,

New York: David McKay.

Baldiris, S., Santos, O. C., Barrera, C., Boticario, J. G.,

Vélez, J. and Fabregat, R. 2008. Integration of

Educational Specifications and Standards to Support

Adaptative Learning Scenarios in ADAPTAPlan,

International Journal of Computer & Applications, 5,

Vol. 1, 88-107

González, R., Valle, A., Núñez, J. C. and González-

Pienda, J. A. 1996. Una aproximación teórica al

concepto de metas académicas y su relación con la

motivación escolar, Psicothema, 8, Vol.1, 45-61

Martín-Albo, J., Núñez, J. L. and Grijalvo, F. 2007. The

Rosenberg Self-Esteem Scale: Translation and

Validation in University Students, The Spanish Journal

of Psycology, 10, Vol. 2, 458-467

Wolpers, M., Najjar, J., Verbert, K. and Duval, E. 2007.

Tracking Actual Usage: the Attention Metadata

Approach, Educational Technology & Society, 10,

Vol. 3, 106-121

Perkins, D. 1995. La escuela inteligente, Ed. Gedisa.

Ng M. K. and Wong J. C. 2002. Clustering categorical

data sets using tabu search techniques, Pattern

Recognition, vol. 35, 2783-2790

Nguyen H. T. and Walker E. A. 2000. A first course in

fuzzy logic, Boca Raton (Florida), Chap-man &

Hall/CRC

A NEW FRAMEWORK FOR THE CONTROL OF LMS IN ITS

289