Abstract:

E-learning platforms automatically log detailed data on the interaction of learners with the platform and its contents; such educational data are source of valuable information that may be exploited to improve the educational process and its outcomes. Educational Data Mining research employs a variety of methods, techniques and tools to extract, process and analyze these data and produce results that will support end users such as students, teachers, researchers and administrators. Efficient and meaningful visualization of the data analysis results is today recognized as an urgent need connected to the increasing data volumes and complexity. Various visualization tools have been developed either for off-line or on-line processing; they may be either embedded in the e-learning platform (internal) or loosely connected to it (external).

The present study is motivated by the observation that visualization of educational data mining results is of prime interest to teachers and class instructors who include e-learning platforms in their teaching scenarios – and yet, teachers and class instructors are rarely involved in the decisions regarding the choice of tools to be installed and employed or in the direct assessment and interpretation of the resulting plots and graphs. This contradiction is attributed (i) to the lack of a ‘guide’ for the teacher that would compare and classify available tools by educational functionalities rather than by technical features, and would thus aid the teacher to make a studied choice according to his/her specific needs, and (ii) to the technical aspects of the installation, parametrization and use of visualization tools that render them unattractive to educators seeking to get a quick, accurate overview of their class performance and behavior. Relevant decisions and parametrization are therefore left to technical staff, e.g. to platform administrators, who are expected to deliver optimal system performance on the basis of qualitatively expressed teachers’ needs.

The present study aims to address these two issues: we carry out a comparative analysis of modern, popular data mining and visualization tools, such as Gephi, WEKA and ProM, as well as of conventional tools (e.g., MS Excel), organized across functional education tasks rather than strictly technical features. This type of comparison constitutes the major contribution of our work, as it is not covered in existing relevant research. The set of tasks adopted comprises visualization of a single variable, joint visualization and correlation of two or more than two variables and 3-D visualization. All tasks are carried out at the individual and at the class level.

The comparative evaluation of these tools is carried out along two sets of criteria:
(i) completeness, spatial organization, information coding, state transition and
(ii) help and orientation features, navigation and browsing features and data set reduction.

Moreover, comparison results are illustrated on real field data from a graduate course offered on the moodle platform in 2020, so that interested teachers may get a quick yet accurate idea on the relative merits of each tools. Results indicate the complementary character of the major existing tools – an outcome that recommends the combined use of more than one tool for optimal performance. Finally, the development of a Teacher Recommendation Tool for use by interested teachers who are not necessarily tech experts is outlined as future research.

Keywords:

e-learning platforms, moodle, educational data mining, visualization, visualization tools, comparative evaluation, Gephi, WEKA, ProM, MS Excel, 3D visualization.