Abstract:

Learning logs stored in LMS (Learning Management System) have the ability to predict learning behaviors and help to give more appropriate feedbacks for learners. In learning log data analysis, we can select various indicators, for example frequency, timing and so on, to predict learning behaviors and performance.

Goda et al. (2015) showed the students of learning habit type scored significantly higher on the TOEIC-IP test than students of the procrastination type, when carrying out an analysis of learning data associated with their e-learning courses. The students of procrastination type tend to procrastinate on a task or action until the last minute. The students of learning habit type form their learning habits. In their case, the students were required to complete the learning materials by the end of the semester. So the deadline was just in the end of the semester. However, in some blended learning cases, students are required to pass some exams until each deadline. When clustering learning behavioral type until each deadline, the clustered results are earned for each exam duration. The periodical clustered results enable to identify students to repeat same learning behavior, including procrastination type.

We provide information literacy classes with blended learning style combined face-to-face lecture and online textbooks and exams on Moodle LMS for 15 weeks. The students were required to complete tasks in each class, to pass more than 2/3 passes of all quizzes outside of the classroom, and a final assessment. So the LMS stored the learning log created by students' activities for their learning.

We analyzed learning logs data associated with online exams of the classes, from October 2018 to February 2019. Exams opened according to class every week, and closed in about two weeks. In the analysis, we firstly divided the exam duration (about two weeks) into four periods. A characteristic vector of each student was made from the number of times, which students accessed to an exam was counted in each period, and the minimum and maximum values ​​of the exam scores in each period were determined in each period. The characteristic vector of a student had 12 elements, access times, the minimum value and maximum value in each period. The characteristic vector of all students were plotted two-dimensionally by the self-organizing map method. The map displayed four groups of similar vector, indicating four learning behavioral types. Then, the hierarchical cluster analysis by Ward method, were classified students into four groups.

The Group1 was a group which access is concentrated just before the deadline. The Group1 was considered to be a group of procrastination type. The Group2 was the group to address one week after their lesson. We consider the Group2 students learned deliberately, but did not learn effectively, because of working on the exam at a later lesson a while. The Group3 and Group4 students were considered to have a learning habit. When comparing the scores of the final assessment between these groups, scoring in the order of Group4, Group3, Group2, Group1 was high.

The students were clustered on the types of the four learning behaviors by our method to analyze the learning log. It was showed that students of learning habit type had higher performance than procrastination type. Using this analysis method, it will be possible to provide appropriate feedback for students except for learning habit type in the future.

Keywords:

Learning types, Higher Education, Learning Analytics, Computer Literacy.