Rankings or leagues or rankings on leagues? - Ranking in fair reference groups

Abstract

There are several well-known rankings of universities and higher education systems. Numerous recent studies question whether it is possible to compare universities and countries of different constitutions. These criticisms stand on solid ground. It is impossible to create a one-dimensional ordering that faithfully compares complex systems such as universities or even higher education systems. We would like to convince the reader that using well-chosen elements of a family of state-of-the-art data mining methods, namely, bi-clustering methods, can provide an informative picture of the relative positions of universities/higher education systems. Bi-clustering methods produce leagues of comparable entities alongside the indicators, which produce a similar grouping of them. Within leagues, partial rankings could be specified and furthermore can serve as a proper basis for benchmarking.