Data mining tools have a recognized status as a part of modern learning environments. Most of the work in data mining in
educational systems contributes to student assessment and course adaptation (for example [2] and [3]). Clustering and classification are mostly used data mining techniques in learning environments because of the nature of the problems. However, learning environments where data mining processes and results would be explicitly visible to the users are rarely reported in the research literature. Rather than that, the data mining takes place in a black box. Systems presented in literature are usually based on approach where a domain expert manually labels data sets and builds models describing the learning activities in advance. This is a time-consuming and error prone process especially in exploratory
learning environments with open-ended problems, because prior definitions of relevant behaviors are necessarily not available for labeling data and training the model [1], and a substantial amount of data may be required to build a model. We have shown in our previous research that the OME produces valuable insight into the progress of the learning activity even on relatively small datasets