Corrigendum to "The unimodal model for the classification of ordinal data" [Neural Networks 21 (2008) 78-79]

In 2008 (Pinto da Costa, Alonso, & Cardoso, 2008) we proposed the unimodal paradigm for ordinal data classification and also a new coefficient to measure the performance of ordinal data classifiers; that is, classifiers for the important and often neglected case where there is an order relation between the classes. Many real life settings exist which involve classifying examples (instances) into classes which have a natural ordering, like for instance econometric modelling; information retrieval and collaborative filtering; stock trading support systems, where one wants to predict, for instance, whether to buy, keep or sell a stock. In fact in a very large number of applications the classes are ordered, although that was rarely taken into account. Recently, the subject has attracted a growing interest (Cruz-Ramírez, HervásMartínez, Sánchez-Monedero, & Gutiérrez, 2011; Hu, Guo, Yu, & Liu, 2010; Pinto da Costa, Sousa, & Cardoso, 2011; Seah, Tsang, & Ong, 2012). We also proposed in Pinto da Costa et al. (2008), apart from the unimodal paradigm, a new coefficient tomeasure the performance of ordinal data classifiers. Since then a growing number of works have also addressed this issue (Baccianella, Esuli, & Sebastiani, 2009; Cardoso & Sousa, 2011; Cruz-Ramírez, Hervás-Martínez, Sánchez-Monedero, & Gutiérrez, 2014; Lee & Liu, 2002; SánchezMonedero, Gutiérrez, Tino, & Hervás-Martínez, 2013; Waegeman, De Baets & Boullart, 2006; Vanbelle & Albert, 2009). The coefficient rint thatwehaveproposed in Pinto daCosta et al. (2008) has aminor error in its final formulationwhichmust be corrected otherwise its