Segmentation of E-Learners through Data Clustering and Classification

In the present paper, descriptive and predictive analytics of e-learner data regarding their behavior in the system is explored with the aim of personalizing e-learning. Descriptive analysis is based on clustering the data (k-means algorithm), identifying the characteristics of learners from different clusters, and summarizing their preferences. Prediction is carried out by applying classification to the defined clusters. Experiments are conducted with various data classification algorithms (k-nearest neighbors algorithm (K-NN), Decision Tree, Naive Bayesian classifier, J48, H2O Deep Learning). The proposed approach is compared with classification based on learning styles determined by learners through questionnaires. The accuracy improvement ranges from 13% to 34%, while the F-measure increases between 23% and 50% depending on the classifier.