Integrating Machine Learning Methods throughout the Temporal Extent of a Web-based Student Model
In this paper we describe the student modeler of an adaptive and intelligent Web-based algebra tutor, which is called Web-EasyMath. The student modeling process consists of two subsequent phases. The first phase concerns the initialization of the student model, whereas the second phase is responsible for updating the student model based on the observed behavior of the student. WebEasyMath makes use of techniques inherited from the area of machine learning in both phases of the student modeling process. In particular, the initialization of the model of a new student is based on a novel combination of stereotypes and the distance weighted k-nearest neighbor algorithm. Furthermore, Web-EasyMath maintains the model of each student using a data ageing mechanism as well as the history of the student's actions.