Clustering Student Mobility Data in 3-way Networks

Abstract

The present contribution aims at introducing a network data reduction method for the analysis of 3-way networks in which classes of nodes of different types are linked. The proposed approach enables simplifying a 3-way network into a weighted two-mode network by considering the statistical concept of joint dependence in a multiway contingency table. Starting from a real application on student mobility data in Italian universities, a 3-way network is defined, where provinces of residence, universities and educational programmes are considered as the three sets of nodes, and occurrences of student exchanges represent the set of links between them. The Infomap community detection algorithm is then chosen for partitioning two-mode networks of students’ cohorts to discover different network patterns.