Visual Clustering of Multidimensional Educational Data From an Intelligent Tutoring System

Abstract

Data mining can be used in the educational field to make teaching more effective, to provide help for the student and teacher, and to analyze data. Data mining can be useful with educational software like intelligent tutoring systems which contain large quantities of data about students and their degree of success, scores, and errors. By analyzing these data, it is possible to obtain pedagogically reliable information as feedback for the teacher. The objectives of this study are to cluster the data available from an Intelligent Tutoring System (ITS) and to visualize the multidimensional data analysis results. In this study, the ITS data, which contain exam results on six different concepts, are clustered using k-means and fuzzy c-means algorithms, and the clustering performance of the two algorithms is compared. Cluster analysis results are visualized using a parallel coordinate system. Clustering and visualizing of the concept-level scores is used to provide meaningful and nontrivial insights into the workings of a course. Such information is useful for the teacher to discover which concepts give students difficulty and which do not. The paper also describes the data mining software developed in this study and the analytical results obtained. (C) 2009 Wiley Periodicals, Inc. Comput Appl Eng Educ 18: 375-382. 2010; Published online in Wiley InterScience (www.interscience.wiley.com); DOI 10.1002/cae.20272