Abstract:

E-learning platform access and usage data are a source of valuable information for the improvement of the educational services offered by the platform and eventually by the educational organization that runs it. E-learning platforms use databases to store detailed log data on the access, the usage and the interaction of individual users with the platform and with the educational material. All this monitoring and registering is transparent to the platform user (student); yet, it produces a wealth of data available for extraction and exploitation by the administrator and the interested researcher. Actions and quantities typically registered in the database include user login times, user access of the various parts of the material, duration of sessions, number and times of keystrokes (clicks, scrolls, page loads), file downloads or file shows, etc. This detailed logging results in large volumes of data; data mining methods are therefore necessary for the selective extraction of useful data from the logs.

An experimental study is described in this paper, where access and usage data from a moodle e-learning platform database are extracted and exploited to answer a series of research questions of educational / paedagogical interest. Questions have to do with students’ practices, strategies and academic performance. The study refers to the data collected through the moodle server of the Department of Electronics Engineering, Technological Education Institute of Piraeus, Greece, during the winter semester of academic year 2015-16. A methodology is proposed and its experimental application is outlined, on a specific e-learning course (Digital Signal Processing Laboratory, 5th semester undergraduate).

Data analysis following data extraction focuses:
(i) on the type of relations among the various quantities extracted from the moodle database (causality, linearity, correlation, etc.),
(ii) the type of relations among the access and usage data and the learning outcomes of the electronic course, and
(iii) the feasibility of prediction of the learning outcomes (students’ performance, in terms of grades) on the basis of access, usage and interaction data.

Results reveal interesting relations among the various quantities, varying from strongly linear to highly non-linear. Clustering methods reveal the interconnection between students’ performance and platform interaction data. Finally, a clearly positive prospect arises from the results as to the possibility to predict students’ performance from platform interaction data.

Keywords:

e-learning platform, moodle, data mining, academic performance, access and usage data, undergraduate curriculum.