Data mining in hybrid learning: Possibility to predict the final exam result

The hybrid learning environment that uses traditional lectures and examinations in conjunction with online learning resources and online assessment tools provides numerous data on students' activities and assessment scores which could be used for constructing a final exam result prediction model. In this paper the data on activities and assessments supported by information and communication technology (ICT) of 302 students enrolled in the first year Physics course of a biomedical university study program have been used to establish the correlations between scores on written midterm exams, scores on web-based formative assessment during seminar teaching, scores on web-based formative assessment during laboratory teaching, scores and time used for online self-assessment test, number of Moodle logins, number of approaches to specific Moodle resources and final exam result. As prediction methods the Principal Component Regression (PCR) and Partial Least Square regression (PLS) have been used, especially due to assumed multi-colinearity of predictive variables and dimension reduction requirement. The model could be useful for students and for teachers who would have the possibility to react and remedy the predicted final exam result if necessary.