25.07.2013 Views

January 2012 Volume 15 Number 1 - Educational Technology ...

January 2012 Volume 15 Number 1 - Educational Technology ...

January 2012 Volume 15 Number 1 - Educational Technology ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

The survey consisted of 30 questions of which some dealt with general information (average grade, year of study)<br />

while the others were about motivation for learning, preferred style of communication, the manner of presenting<br />

subject matter, managing time and team work.<br />

II phase – Introductory course<br />

In order to obtain more data about students’ learning styles, we organized an introductory course that lasted for one<br />

week, involving all types of activities, resources, and materials. The course was created in Moodle LMS. It dealt with<br />

different aspects of e-business: e-commerce, e-government, customer relationship management, Internet marketing.<br />

The course was not adapted in any way, just like the other courses in Moodle. In order to complete the course,<br />

students had to pass a test designed to assess knowledge acquired from different types of materials and activities: text<br />

and multimedia materials, web pages, as well as activities available in Moodle (assignments, lessons, quizzes, forum<br />

discussions, etc). The test results were to be used to determine students’ learning styles.<br />

3. Exploring collected data<br />

All the data collected in the two described phases were integrated into a single table in such a way that each question<br />

from the questionnaire and from the test represented one column. The rows represented answers of the students. This<br />

table was suitable for further analysis and data mining. Upon integration, the data were transformed and reduced.<br />

The number of options in the answers were transformed and reduced to three, and the missing data were replaced<br />

with mean values.<br />

4. Classifying students<br />

Data mining is one of the business intelligence techniques and can be defined as the nontrivial extraction of implicit,<br />

previously unknown and potentially useful information from large data sets or databases 0. Although personalized<br />

recommendation approaches that use data mining techniques have been first proposed and applied in e-commerce for<br />

product purchase, different data mining techniques were also applied within the e-learning recommender systems<br />

000.<br />

By using data mining tools and techniques it is possible to conduct an intelligent analysis of large quantities of data<br />

stored in a database. Data mining can be used both as a means for predicting unknown or future values of the<br />

attributes of interest and for describing embedded patterns that will contribute to generating the best possible<br />

personalized e-learning models 0 00.<br />

Clustering, as a data mining technique, was applied in building data mining model on the integrated data set prepared<br />

in the previous phase. Clustering algorithm finds natural groupings among data related to sets of input attributes, so<br />

that attributes within one group (cluster) have approximately the same values, while notable differences exist<br />

between groups (clusters) 0. It could be asserted that the main aim of clustering is to discover hidden values and<br />

variables, upon which data can be precisely arranged. Clustering was carried out using the SQL Server Analyses<br />

services 0, a Microsoft clustering algorithm based on the K-mean algorithm 0 0.<br />

A set of guiding questions were formulated in order to lead the process of creating data mining structures in the right<br />

direction:<br />

What is the most appropriate number of clusters?<br />

What are the main characteristics within the clusters and what are the differences between them?<br />

Which input variable, i.e. learning style, has a dominant influence in grouping the students?<br />

Is the mining model created suitable for making forecasts?<br />

Which content should be delivered to students from individual clusters?<br />

The answers to these questions are given in the text that follows.<br />

An experiment was conducted for the cases of two and three clusters. By processing and mining available data, it<br />

was determined that the results showed almost the same level of accuracy, regardless of whether the students were<br />

divided into two or three clusters. However, in the latter case, the outcomes were more consistent, logical and of<br />

higher quality. Therefore, the outputs and the conclusions drawn from the experiment with three groups will be<br />

presented here.<br />

330

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!