SMOTE for regression

• Tipo: Artigo em Livro de Atas de Conferência Internacional
• Data: 2013

Autores

Torgo, L (Autor) FCUP; Ribeiro, RP (Autor) FCUP; Pfahringer, B (Autor) Outra; Branco, P (Autor) Outra

Ata de Conferência Internacional

Título: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

Páginas: 378-389

16th Portuguese Conference on Artificial Intelligence, EPIA 2013

Angra do Heroismo, Azores, 9 September 2013 through 12 September 2013

Indexação

ISI Web of Knowledge

Scopus - 166 Citações

Outras Informações

ID Authenticus: P-008-EG1

DOI: 10.1007/978-3-642-40669-0_33

Abstract (EN): Several real world prediction problems involve forecasting rare values of a target variable. When this variable is nominal we have a problem of class imbalance that was already studied thoroughly within machine learning. For regression tasks, where the target variable is continuous, few works exist addressing this type of problem. Still, important application areas involve forecasting rare extreme values of a continuous target variable. This paper describes a contribution to this type of tasks. Namely, we propose to address such tasks by sampling approaches. These approaches change the distribution of the given training data set to decrease the problem of imbalance between the rare target cases and the most frequent ones. We present a modification of the well-known Smote algorithm that allows its use on these regression tasks. In an extensive set of experiments we provide empirical evidence for the superiority of our proposals for these particular regression tasks. The proposed SmoteR method can be used with any existing regression algorithm turning it into a general tool for addressing problems of forecasting rare extreme values of a continuous target variable. © 2013 Springer-Verlag.

Idioma: Inglês

Tipo (Avaliação Docente): Científica

Nº de páginas: 12