Paper Info
Title
Is Students' Activity in LMS Persistent?
Abstract
Int roduction. The most common method of blending the Internet in higher education today is by implementing Web-supported instruction, in which traditional face-to-face courses have auxiliary materials, usually using Learning Management Systems (LMS), e.g. WebCT, Moodle. Research of LMS in higher education has barely involved the examination of the individual's behavior over the learning period. Furthermore, although a large body of research exists regarding persistence in fully online learning configurations, only little was studied regarding the online persistence in Web-supported configurations. When empirically examining usage of Web learning environments, it has been noticed that two phenomenon are repeatedly occurring regarding volume and trends of activity: a) Many are little active, while some are extensively active; b) Overall decrease in visiting (usually with some spikes of access immediately before exams, assignment submission deadlines, or any other important events during the course) (1-4). This study aims on identifying individuals' over-time patterns of online activity in Web-supported courses, both by volume and trends of activity. As our examination of patterns of persistence crosses courses, it might also promote the revealing of differences between courses regarding students' persistence within them. Population. Log files of 58 Moodle one-semester-course websites offered by Tel Aviv University (TAU) in the academic year 2008/9 were analyzed (a full sample of the Moodle-supported courses; only logged activity from during the calendared term period were taken). Moodle's log files consist of actions taken within the course websites' modules, including: text pages, resources, forums, and users. Actions might be: viewing, adding, updating, or deleting. In total, 163,685 records of 1189 students were logged, and there were 1897 student enrollments which served as the basic analysis units (interdependence in the population was found to be insignificant). Variables and Process. Five measures were calculated to describe students' activity in volume (Cumulative Activity, Total Activity) and trend (First Tertile Proportion, Second Tertile Proportion, Activity per Day). The main procedure involves the application of a Decision Tree algorithm on the trends-related variables, for finding patterns of persistence in students' behavior, and for defining rules of belonging to these patterns. We choose the variable Activity per Day as the independent variable the prediction of which should be given by the tree, and the two other variables - i.e., First/Second Tertile Proportion - as the variables according to which the tree will be constructed. CHAID method was used as an attribute selection measure based on the statistical chi-square test for independence, with a significance level of 0.05 for splitting nodes and merging categories, and a 10-fold cross validation. Results and Discussion. Analyzing Total Activity, it was re-demonstrated that most of the students present low activity, while only a little are very active. Regarding the
Year
Venue
Keywords
2010
EDM
decision tree,higher education,chi square test,cumulant,attribute selection,cross validation
Field
DocType
Citations 
Online learning,Learning Management,Computer science,Artificial intelligence,Phenomenon,Machine learning,Higher education,Web learning,The Internet
Conference
1
PageRank 
References 
Authors
0.35
5
2
Name
Order
Citations
PageRank
Arnon Hershkovitz110613.39
Rafi Nachmias235038.59