Mining Web-based Educational Systems to Predict Student Learning Achievements

Abstract

Educational Data Mining (EDM) is getting great importance as a new interdisciplinary research field related to some other areas. It is directly connected with Web-based Educational Systems (WBES) and Data Mining (DM, a fundamental part of Knowledge Discovery in Databases). The former defines the context: WBES store and manage huge amounts of data. Such data are increasingly growing and they contain hidden knowledge that could be very useful to the users (both teachers and students). It is desirable to identify such knowledge in the form of models, patterns or any other representation schema that allows a better exploitation of the system. The latter reveals itself as the tool to achieve such discovering. Data mining must afford very complex and different situations to reach quality solutions. Therefore, data mining is a research field where many advances are being done to accommodate and solve emerging problems. For this purpose, many techniques are usually considered. In this paper we study how data mining can be used to induce student models from the data acquired by a specific Web-based tool for adaptive testing, called SIETTE. Concretely we have used top down induction decision trees algorithms to extract the patterns because these models, decision trees, are easily understandable. In addition, the conducted validation processes have assured high quality models.