Universiti Teknologi Malaysia Institutional Repository

Granular mining approach for identifying student's learning style in e-learning environment

Ahmad, Nor Bahiah (2012) Granular mining approach for identifying student's learning style in e-learning environment. PhD thesis, Universiti Teknologi Malaysia, Faculty of Computer Science and Information System.


Official URL: http://dms.library.utm.my:8080/vital/access/manage...


Pattern multiplicity of interaction in e-learning can be intelligently examined to diagnose students’ learning style. This is important since a student’s behaviour while learning online is among the significant parameters for adaptation in elearning system. Currently, Felder Silverman (FS) is a common learning style model that is frequently used by many researchers. There are four learning style dimensions in FS model and most researches need to develop four classifiers to map the characteristics into the dimensions. Such approach is quite tedious in terms of data pre-processing and it also time consuming when it comes to classification. Therefore, this study improves the previous work by mapping the students’ characteristics into Integrated Felder Silverman (IFS) learning style, by combining the four learning dimensions in FS model into sixteen learning styles. The most crucial problem for IFS model is the difficulties in identifying the significant pattern for the classifier that has high dimension and large number of classes. In this study, fifteen features have been identified as the granule learning features for learning style recognition based on the analysis resulting from questionnaire and log data. The granularity of the learning features is efficiently implemented using Rough Set Boolean Reasoning and Genetic Algorithm. However, Rough Set generates huge rules that are redundant and irrelevant. Hence, these rules need to be incrementally pruned to extract the most significant one. The rules are pruned by evaluating the rules support, the rules length and the rules coverage. The experiment shows that with only 12 per cents rules left, the classification accuracy is still significant and the rule coverage is also high. Comparative analysis of the performance between IFS classifier and the conventional four classifiers shows that the proposed IFS gives higher classification accuracy and rule coverage in identifying student’s learning style.

Item Type:Thesis (PhD)
Additional Information:Thesis (Ph.D (Sains Komputer)) - Universiti Teknologi Malaysia, 2012; Supervisor : Prof. Dr. Siti Mariyam Shamsuddin
Uncontrolled Keywords:learning strategies, learning, psychology, computer-assisted instruction, internet in education
Subjects:L Education > L Education (General)
L Education > LB Theory and practice of education
Divisions:Computer Science and Information System
ID Code:33351
Deposited By: Kamariah Mohamed Jong
Deposited On:01 Nov 2013 02:45
Last Modified:20 Sep 2017 00:44

Repository Staff Only: item control page