Improving the Classification of Study-related Data throughSocial Network Analysis

The Information System of Masaryk University (IS MU) hosts applications utilized for managing study-related records, e-learning tools and those facilitating communication inside the University. This paper is concerned with improvement of results obtained with Excalibur, a tool for mining study-related data, when linked data have been added. These data describe social dependencies gathered from e-mail and discussion boards conversation. We first describe results based on the original (non-linked) data that are periodically saved into Excalibur data warehouse. Then focus on extraction of social dependencies namely relations and communication among students. We describe a method for feature extraction from the social dependencies. New features were explored by social network analysis and visualization tool Pajek and added to the original data. We show that such enriched data allows to significantly improve results obtained with data mining methods. We demonstrate this general technique on different tasks that concern classification of successful/non-successful students at Faculty of Informatics MU.