April 2012 Volume 15 Number 2 - Educational Technology & Society

BioLayout uses a modified version of the Fuchterman and Rheingold graph layout algorithm. It is used to produce an
“aesthetically pleasing” layout of complex graphs (Enright et al., 2002). Based on this algorithm two users that are
similar, meaning they have a connection with a high number of web pages that they both have visited, will end up
closer in the final graph than two users which are weakly similar. Highly connected groups of similar users will form
a tight cluster in the final graph.
It must also be noted that the MCL algorithm can be run under the BioLayout environment directly using the graph
currently loaded in BioLayout and the resulting clusters are automatically coloured to help distinguish them.
In order to annotate each cluster, in a clutter-free way, we use the UserID and the CourseIDs he/she has visited to
label each node of the cluster. This annotation schema allows the end-user to better understand the information in
each cluster. Using the search capabilities of BioLayout the user can also easily find specific Courses in the graph
and the users and groups related to them.
Experimental scenario
In order to evaluate the proposed methodology and assess its usefulness as a data mining tool for LMS systems, we
used a dataset from a real e-learning environment.
Dataset description
The dataset was collected from the Technological Education institute (TEI) of Kavala that uses the Open eClass elearning
platform (GUNet, 2009). The data are from the spring semester of 2009 from the Department of Information
management and involve 1199 students and 39 different courses. The data are in ASCII form and are obtained from
the Apache server log file.
Preprocessing
The log file produced in the previous step is filtered, so it includes only the fields: (i) courseID, (ii) sessionID, and
(iii) page Uniform Resource Locator (URL), as described in detail in the methodology section. Table 2 presents the
results for 20 courses. The first 10 courses are with the highest Final Score, the next 5 courses are the ones ranked in
the middle and the last 5 are the lowest ranked. We wanted to test our metrics, in best, average and worst cases from
a usage point of view.
Table 2. Processed e-learning data for 20 Courses
Course ID Sessions Pages Unique pages UPCS Interest Enrichment Quality index Final score
IMD105 91 297 11 216 0,694 0,963 0,828 178,91
IMD35 87 338 8 179 0,743 0,976 0,859 153,84
IMD132 152 230 7 184 0,339 0,970 0,654 120,40
IMD36 72 217 7 134 0,668 0,968 0,818 109,61
IMD129 75 209 6 131 0,641 0,971 0,806 105,61
IMD125 93 164 6 134 0,433 0,963 0,698 93,55
IMD41 98 185 8 129 0,470 0,957 0,714 92,04
IMD66 56 144 9 107 0,611 0,938 0,774 82,85
IMD17 53 206 11 89 0,743 0,947 0,845 75,17
IMD111 33 142 9 79 0,768 0,937 0,852 67,32
IMD115 10 73 12 42 0,863 0,836 0,849 35,67
IMD9 26 105 12 42 0,752 0,886 0,819 34,40
IMD120 38 80 3 46 0,525 0,963 0,744 34,21
IMD112 30 62 6 46 0,516 0,903 0,710 32,65
160