Sabine Rathmayer, Hans Pongratz (Hrsg.): Proceedings of DeLFI Workshops 2015 co-located with 13th e-Learning Conference of the German Computer Society (DeLFI 2015) München, Germany, September 1, 2015 101 Educational Data Mining / Learning Analytics: Methods, Tasks and Current Trends Agathe Merceron 1 Abstract: The 1 st international conference on “Educational Data Mining” (EDM) took place in Montreal in 2008 while the 1 st international conference on “Learning Analytics and Knowledge” (LAK) took place in Banff in 2011. Since then the fields have grown and established themselves with an annual international conference, a journal and an association each, and gradually increase their overlapping. This paper begins with some considerations on big data in education. Then the principal analysis methods used with educational data are reviewed and are illustrated with some of the tasks they solve. Current emerging trends are presented. Analysis of educational data on a routine basis to understand learning and teaching better and to improve them is not a reality yet. The paper concludes with challenges on the way. Keywords: Educational data mining, learning analytics, prediction, clustering, relationship min- ing, distillation of data for human judgment, discovery with models, multi modal analysis, multi- level analysis, natural language processing, privacy, data scientist. 1 Introduction “Big Data in Education” was the name of a MOOC offered on Coursera in 2013 by Ryan Baker. What means big data in education? To answer this question I consider different sources of educational data following the categorization of [RV 10]. Schools and univer- sities use information systems to manage their students. Take the case of a small- medi- um European university with 12 000 students and let us focus on the marks. Assuming that each student is enrolled in 6 courses, each semester the administration records 60 000 new marks (including the null value when students are absent). Many universities and schools use a Learning Management System (LMS) to run their courses. Let us take an example of a small course, not a MOOC, taught for 60 students on 12 weeks with one single forum, and a set of slides and one quiz per week. LMSs record students’ interactions, in particular when students click on a resource, write or read in the forum. Assume that each student clicks on average twice each week on the set of slides and the quiz, and 3 times on the forum in the semester. This gives 3060 interactions that are stored for one course during one semester. Let us suppose that the small-medium university from above has 40 degree-programs with 15 courses each. This gives 1 836 000 interactions stored by the LMS each semester. 1 Beuth Hochschue für Technik, Fachbereich Medieninformatik, Luxemburgerstrasse 19, 13353 Berlin, merce- ron@beuth-hochschule.de