Abstract:

Studies show that learning approaches (deep or surface) used by the students have an impact on their academic performance. Moreover students using surface learning approaches tend to have lower academic performance. Therefore, it is an important task for educators to detect the students having surface approach and to help them to use deep learning strategies. A set of self-reported instruments is proposed to measure students' approaches to learning. However automatic detection of students’ learning approach based on their actions in online learning environment will increase the effectiveness of the interventions. So, the present study aimed to develop a prediction model to estimate students' learning approaches through the analysis of their interactions with Moodle. The participants of the study were 60 third year undergraduate students enrolled in a Relational Database Management System course. Learning approaches of students were measured at the end of the course by the Revised Two Factor Study Process Questionnaire (R-SPQ-2F). The R-SPQ-2F consists of 20 items with ten items measuring surface approach to learning and ten items measuring deep approach to learning. Students are divided into groups according to their scores from the subscales of the R-SPQ-2F using the clustering method. Then a classification analysis was applied for predicting students' learning approach with the data obtained from the Moodle logs. Cluster analysis performed using the RapidMiner software with K-medoids clustering algorithm, the classification analysis was performed using Orange data mining software with K-Nearest Neighbors (k-NN) algorithm. Classification analysis results were generalized using the 10k cross-validation method. Two groups of students were identified according to the results of the cluster analysis. They are labeled with regard to cluster centers. The first one is Deep Learners (n = 32) and the second one is Surface Learners (n = 28). When academic success of the students in different clusters are examined it is seen that 13 of the 17 students who failed the course were surface learners (76.5%), on the other hand 30 of the 45 students who passed the course were deep learners (66.7%). When cross-validated classification analysis results are analyzed it is seen that classification model obtained from k-NN algorithm accurately classified 29 of the 34 deep learners (85.3%), and 25 of to 28 surface learners (89.3%). In other words, the obtained classification model showed that students learning approaches can be predicted based on their Moodle activity. These activities involve participating in discussions, performing assignments, viewing resources, and etc.

Keywords:

Educational data mining, learning approaches, prediction, R-SPQ-2F, classification.