The aim of this study is to develop a classification model that predicts learners' approaches (deep, surface) towards the Database Management System course based on their interaction data in an online learning environment. Therefore, first, participants' approaches to learning are measured with The Revised Two Factor Study Process Questionnaire (R-SPQ-2F) and through cluster analysis the participants are divided into two groups in accordance with their scores from the sub-dimensions of the scale. These groups are labeled as Deep Learner (n = 34) and Surface Learner (n = 28) considering the cluster centers. Following that, a prediction model is developed to predict the students' approaches to learning upon activities gathered from Moodle logs by using classification analysis. Cross validated classification analysis results show that 29 out of 34 deep learners (85.3%) and 25 out of 28 surface learners, together which equals to 86.9 percent of all the participants, have classified correctly by the classification model generated by k-NN algorithm. In other words, it has been found out that learners' such activity data as submitting assignments, participating in discussions and logging in Moodle could predict their learning approaches remarkably. Besides, academic achievements of the learners from different clusters are also investigated within the scope of this study.