In this paper, we present a data mining approach for analysing students' clickstream data logged by an e-learning platform and we propose a machine learning procedure to predict course completion of students. For this, we used data from a short MOOC course which was motivated by the teachers of elementary schools. We show that machine learning approaches can accurately predict the course outcome based on clickstream data and also highlight patterns in data which provide useful insights to MOOC developers.

Nowadays online presentations and educational videos are frequently integrated into various-learning environments and applications, such as MOOCs, global sets of conferences or video-sharing websites. This paper presents the findings of a comparison of online presentations (educational videos) and offline presentations.The total number of student participants in this research was 191, mostly primary and secondary school students from Serbia as well as Hungary, studying both online and offline learning environments within the framework of the course Conscious and safe internet usage. The impact of offline and online presentations was investigated using both pre-and post-presentation questionnaires. Statistical analysis was used to measure the impact of offline and online presentations, in addition to other factors determining student achievements.

In today's world virtual online educational platforms emerge literally on daily bases and many offer MOOC-based courses. With the appearance of MOOC, educational platforms have gained an additional boost, a new aspect in their evolutionary process, which has opened a new field of research thanking to the extraction of logging information within the frames of data mining. It has become clear that educators will be able to tailor their courses by merging the two previously mentioned fields and by carrying out MOOC-based data mining, targeting pedagogical aspects. This field of research seems promising and important, thus a faculty at the University of Szeged has created its own MOOC educational platform which has been set to facilitate data mining by implementing a wide range of logging algorithms. The data would be processed through a complex Artificial Intelligence program, which, in the short term, could reveal new and exciting pedagogical findings, while in the long run, the supervisors could put together a platform that would help and notify educators about relevant information. It would become possible to create adaptive educational materials, as well. This work aims at clarifying how such platforms function and what the steps of data collection and evaluation are. I.

Online education has gained a wide popularity in today's global information boom. Prominent universities offer more and more online courses with modern audio-visual content, which have become available for almost everyone. The courses can be completed self-paced allowing for much more flexibility. Such a learning approach has already reformed higher education and seeks ways to penetrate into the realm of the secondary and primary education. Our team conducted a research in the higher classes of different primary schools in the Province of Vojvodina, Serbia. Participants of the Hungarian minority took part in a course named "Conscious and Safe Internet Use" and obtained a valuable knowledge with the help of videos and optional course materials. Student activities were all recorded with the help of a special self-developed logging software for later processing. Data were processed by statistical and data mining methods. According to the values of correlation coefficients and R Square Statistical value it can be stated, that those participants who scored well at the pre-test stage expectedly had better results during the final testing stage. The number of video views, the age of students and the size of their hometown had also influenced the outcome of the tests. The detailed findings are presented in this paper under the Results chapter.