Logo TU Ilmenau




Automatic Classifiers for Formative Assessment

M.Sc. Wassim Mahfouz
Dr.-Ing. Heinz- Dietrich Wuttke
Date of publication
This Research to Practice Full Paper is driven by the question: How can a teacher of a large class be effectively supported for formative assessment? The paper suggests a framework for automatic classifiers / predictive models and its data integration tool for teachers. To design an input data set for automatic classifier, the tool enables the teachers to integrate data extracted from paper-based exams; computer assisted formative assessments and LMSs, with learning disposition data, collected by applying the Buckingham Shum and Deakin Crick’s theoretical framework for dispositional learning analytics. The suggested framework and its tool are tested with the real assessment data of 129 students collected during conducting the Computer Organization (hereafter CO) course. The results of this CO test scenario show how teachers can interpret the outcomes of the automatic classifiers as decision-support recommendations to improve the planning for formative assessment. As an example, it is presented and discussed how the CO teacher can improve his/her strategies in formative assessment for different students groups (at-risk students, medium students, good students and excellent students). The paper shows also how reports for classifiers accuracy comparison can be produced and understood by teachers. Related works are discussed to show the differences and benefits of the presented framework. Main advantages are: the possibility to use its automatic classification algorithms instead of the statistical regression algorithms and the possibility to use its integration tool to integrate data collected from applying a theoretical framework with data from e-learning /e-assessment as input data sets for automatic classifiers.
External link