Other Applications: Sequential Dependence Modelling and Data Mining - On Statistical Pattern Recognition in Independent Component Analysis Mixture Modelling - page 158

Epoch number

2

(a)

1.5

1

50 100 150 200 250 300 350 400

2

(b)

1.5

1

50 100 150 200 250 300 350 400

2

(c)

1.5

1

50 100 150 200 250 300 350 400

Fig. 7.3 Hypnograms corresponding to patient 1. a Expert hypnogram. b SICAMM hypnogram.
c ICAMM hypnogram

7.2 Webmining Application for Detecting Learning Styles
in Virtual Education

7.2.1 Introduction

Pattern analysis or data mining from web log data (webmining) is a new research
area that attempts to understand the flow of information on the web by means of
automated techniques for knowledge search 0 [ 13 ]. This area has a wide range of
emergent applications including e-learning, e-commerce, automated information
assistants, and many applications that operate through the web [ 14 ]. One example
of webmining is the classification of web pages based on understanding the textual
content of emails based on hierarchical probabilistic clustering [ 15 ].
Specifically, there is an increasing interest in webmining from e-learning data.
Some examples are: predicting the drop-out rate using demographic (sex, age,
marital status, etc.) and course data [ 16 ]; predicting the course grades using the
processing success rate, success at first try, number of attempts, time spent on the
problem, etc. [ 17 ]; combining several weak classifiers by boosting in order to
predict the final grades [ 18 ]. Recently, new holistic webmining approaches have
undertaken the extraction of learning styles from web navigational behaviour
outlined by students [ 19 - 21 ].
The approaches used to extract patterns of the web log data have used different
statistical classification and machine learning techniques including ICA. As
explained in Chap. 2 , many of the standard ICA algorithms are based on a