Domain specific interactive data mining

Abstract

Finding patterns in data collected from interactions with an educational system is only useful if the patterns can be meaningfully interpreted in the context of the student- system interaction. To further increase the chance of finding such meaningful patterns, we extend the mining process with domain and problem specific representations and the pattern detection expertise of qualified users. The user, that is, the researcher looking for patterns, is not just evaluating the result of an automatic data mining process, but is actively involved in the design of new representation and the search for patterns. This approach is used in addition to more traditional methods and has resulted in a deeper understanding of our data.