Weighting efficient Accuracy and Minimum Sensitivity for evolving multi-class classifiers

Áreas de investigación:

• Methodology - Machine learning

Año:

2011

Tipo de publicación:

Artículo

Palabras clave:

Artificial Neural Networks, Extreme learning machine, Evolutionary ELM, Multi-class, imbalanced datasets, Accuracy, Sensitivity, Differential Evolution

Autores:

• Sánchez-Monedero, Javier
• Gutiérrez, Pedro Antonio
• Fernandez-Navarro, Francisco
• Hervás-Martínez, César

Journal:

Neural Processing Letters

Volumen:

34

Número:

2

Páginas:

1370-4621

BibTex:

@article{Sanchez2011NPL, author = "Javier S{\'a}nchez-Monedero and Pedro Antonio Guti{\'e}rrez and Francisco Fernandez-Navarro and C{\'e}sar Herv{\'a}s-Mart{\'i}nez", abstract = " Recently, a multi-objective Sensitivity–Accuracy based methodology has been proposed for building classifiers for multi-class problems. This technique is especially suit- able for imbalanced and multi-class datasets. Moreover, the high computational cost of multi- objective approaches is well known so more efficient alternatives must be explored. This paper presents an efficient alternative to the Pareto based solution when considering both Minimum Sensitivity and Accuracy in multi-class classifiers. Alternatives are implemented by extending the Evolutionary Extreme Learning Machine algorithm for training artificial neural networks. Experiments were performed to select the best option after considering alternative proposals and related methods. Based on the experiments, this methodology is competitive in Accuracy, Minimum Sensitivity and efficiency. ", awards = "JCR (2011): 0.750 (category COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE, position 76/111 Q3)", comments = "JCR (2011): 0.750 (category COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE, position 76/111 Q3)", doi = "10.1007/s11063-011-9186-9", journal = "Neural Processing Letters", keywords = "Artificial Neural Networks, Extreme learning machine, Evolutionary ELM, Multi-class, imbalanced datasets, Accuracy, Sensitivity, Differential Evolution", note = "JCR (2011): 0.750 (category COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE, position 76/111 Q3)", number = "2", pages = "1370-4621", title = "{W}eighting efficient {A}ccuracy and {M}inimum {S}ensitivity for evolving multi-class classifiers", url = "http://dx.doi.org/10.1007/s11063-011-9186-9", volume = "34", year = "2011", }

Nota:

JCR (2011): 0.750 (category COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE, position 76/111 Q3)

Abstract:

Recently, a multi-objective Sensitivity–Accuracy based methodology has been proposed for building classifiers for multi-class problems. This technique is especially suit- able for imbalanced and multi-class datasets. Moreover, the high computational cost of multi- objective approaches is well known so more efficient alternatives must be explored. This paper presents an efficient alternative to the Pareto based solution when considering both Minimum Sensitivity and Accuracy in multi-class classifiers. Alternatives are implemented by extending the Evolutionary Extreme Learning Machine algorithm for training artificial neural networks. Experiments were performed to select the best option after considering alternative proposals and related methods. Based on the experiments, this methodology is competitive in Accuracy, Minimum Sensitivity and efficiency.

Comentarios:

JCR (2011): 0.750 (category COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE, position 76/111 Q3)