Other Applications: Sequential Dependence Modelling and Data Mining - On Statistical Pattern Recognition in Independent Component Analysis Mixture Modelling - page 160

Fig. 7.5 E-learning facilities outline at UPA

Table 7.3 Dimensions of learning and teaching styles (Felder's model)
Preferred learning style Corresponding teaching style

1 Sensory-intuitive Perception Concrete-abstract Content

2 Visual-auditory Input Visual-verbal Presentation

3 Inductive-deductive Organization Inductive-deductive Organization

4 Active-reflective Processing Active-passive Student participation

5 Sequential-global Understanding Sequential-global Perspective

7.2.2 ICA Statement of the Problem

In order to solve the problem of detecting learning styles in e-learning, we defined
the following ICA generative model for the observed data.

x ¼ E d ð 7 : 5 Þ

We assume that the underlying independent sources that generate the web log data
are dimensions of the learning styles of the students, and we observe x linear combi-
nations of those styles through the use of the facilities at the virtual campus. Then,
d i ; ð i ¼ 1 ; ... ; 5 learning style dimension Þ corresponds to the ''perception'', ''input'',
''organization'', ''processing'', and ''understanding'' dimensions (see Table 7.3 );
and the mixture matrix E provides the relation between the e-learning style dimen-
sions and the e-learning event activities, e ij ; ð i ¼ 1 ; ... ; 5 learning style dimension Þ;
ð j ¼ 1 ; ... ; 10 e learning activity Þ .
Some preprocessing was done; a datawarehouse was created from the historical
(2001-2005) web data of the UPA. The total number of events for the analyzed
period was 2,391,003 (see the list of e-learning event activities in Sect. 7.2.1 ). The
projection of this event table on the student codes was a table with 8,909 records.