Citation-based journal ranks: The use of fuzzy measures

This article considers a real-world application of the Choquet integral-based classifier to the problem of ranking of scientific journals. We attempted to predict rankings given by panels of experts, by using available citation indices. We compare the Choquet integral fit by the least absolute deviation criterion as a classifier against other ordinal classification methods, before interpreting the fuzzy measure weights and indices. Our findings show that the journal rankings data set is difficult to model accurately due to inconsistencies and lack of monotonicity, but that the Choquet integral still performs well as a classifier. (C) 2010 Elsevier B.V. All rights reserved.