Predicting students' learning achievement by using online learning patterns in blended learning environments: Comparison of two cases on linear and non-linear model

A variety of studies using educational data traced from LMS has been conducted to predict students’ performance. However, because of the complexity in its implementation, it is still challenging to predict students’ learning achievement in blended learning environment. As an exploratory study, we selected two types of blended learning classes and compared their prediction models. While the first blended learning class which involves online discussion-based learning revealed a linear regression model, the second case, which was a lecture based blended learning class providing regular base online lecture notes in Moodle, did not present a linear regression model. After that, to examine the important variables of each class, RF (random forest) method was utilized. The results indicated different important variables in two cases. We concluded that the prediction models and data-mining technique should be based on the considerations of diverse pedagogical characteristics in blended learning.