A scalable pairwise class interaction framework for multidimensional classification

Jacinto Arias, Jose A. Gámez, Thomas Dyhre Nielsen, José M. Puerta

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

4 Citationer (Scopus)

Resumé


We present a general framework for multidimensional classification that cap- tures the pairwise interactions between class variables. The pairwise class inter- actions are encoded using a collection of base classifiers (Phase 1), for which the class predictions are combined in a Markov random field that is subsequently used for multidimensional inference (Phase 2); thus, the framework can be positioned between multilabel Bayesian classifiers and label transformation-based approaches. Our proposal leads to a general framework supporting a wide range of base classifiers in the first phase as well as different inference methods in the second phase. We describe the basic framework and its main properties, as well as strategies for ensuring the scalability of the framework. We include a detailed experimental evaluation based on a range of publicly available databases. Here we analyze the overall performance of the framework and we test the behavior of the different scalability strategies proposed. A comparison with other state-of-the-art multidimensional classifiers show that the proposed framework either outperforms or is competitive with the tested straw-men methods.
OriginalsprogEngelsk
TidsskriftInternational Journal of Approximate Reasoning
Vol/bind68
Sider (fra-til)194–210
ISSN0888-613X
DOI
StatusUdgivet - 2016

Fingerprint

Pairwise
Classifiers
Interaction
Scalability
Classifier
Labels
Bayesian Classifier
Framework
Class
Experimental Evaluation
Range of data
Random Field
Prediction

Citer dette

@article{ef3d6ce5790149ae9b3ea4dd01a9e0b1,
title = "A scalable pairwise class interaction framework for multidimensional classification",
abstract = "We present a general framework for multidimensional classification that cap- tures the pairwise interactions between class variables. The pairwise class inter- actions are encoded using a collection of base classifiers (Phase 1), for which the class predictions are combined in a Markov random field that is subsequently used for multidimensional inference (Phase 2); thus, the framework can be positioned between multilabel Bayesian classifiers and label transformation-based approaches. Our proposal leads to a general framework supporting a wide range of base classifiers in the first phase as well as different inference methods in the second phase. We describe the basic framework and its main properties, as well as strategies for ensuring the scalability of the framework. We include a detailed experimental evaluation based on a range of publicly available databases. Here we analyze the overall performance of the framework and we test the behavior of the different scalability strategies proposed. A comparison with other state-of-the-art multidimensional classifiers show that the proposed framework either outperforms or is competitive with the tested straw-men methods.",
author = "Jacinto Arias and G{\'a}mez, {Jose A.} and Nielsen, {Thomas Dyhre} and Puerta, {Jos{\'e} M.}",
year = "2016",
doi = "10.1016/j.ijar.2015.07.007",
language = "English",
volume = "68",
pages = "194–210",
journal = "International Journal of Approximate Reasoning",
issn = "0888-613X",
publisher = "Elsevier",

}

A scalable pairwise class interaction framework for multidimensional classification. / Arias, Jacinto; Gámez, Jose A.; Nielsen, Thomas Dyhre; Puerta, José M.

I: International Journal of Approximate Reasoning, Bind 68, 2016, s. 194–210.

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

TY - JOUR

T1 - A scalable pairwise class interaction framework for multidimensional classification

AU - Arias, Jacinto

AU - Gámez, Jose A.

AU - Nielsen, Thomas Dyhre

AU - Puerta, José M.

PY - 2016

Y1 - 2016

N2 - We present a general framework for multidimensional classification that cap- tures the pairwise interactions between class variables. The pairwise class inter- actions are encoded using a collection of base classifiers (Phase 1), for which the class predictions are combined in a Markov random field that is subsequently used for multidimensional inference (Phase 2); thus, the framework can be positioned between multilabel Bayesian classifiers and label transformation-based approaches. Our proposal leads to a general framework supporting a wide range of base classifiers in the first phase as well as different inference methods in the second phase. We describe the basic framework and its main properties, as well as strategies for ensuring the scalability of the framework. We include a detailed experimental evaluation based on a range of publicly available databases. Here we analyze the overall performance of the framework and we test the behavior of the different scalability strategies proposed. A comparison with other state-of-the-art multidimensional classifiers show that the proposed framework either outperforms or is competitive with the tested straw-men methods.

AB - We present a general framework for multidimensional classification that cap- tures the pairwise interactions between class variables. The pairwise class inter- actions are encoded using a collection of base classifiers (Phase 1), for which the class predictions are combined in a Markov random field that is subsequently used for multidimensional inference (Phase 2); thus, the framework can be positioned between multilabel Bayesian classifiers and label transformation-based approaches. Our proposal leads to a general framework supporting a wide range of base classifiers in the first phase as well as different inference methods in the second phase. We describe the basic framework and its main properties, as well as strategies for ensuring the scalability of the framework. We include a detailed experimental evaluation based on a range of publicly available databases. Here we analyze the overall performance of the framework and we test the behavior of the different scalability strategies proposed. A comparison with other state-of-the-art multidimensional classifiers show that the proposed framework either outperforms or is competitive with the tested straw-men methods.

U2 - 10.1016/j.ijar.2015.07.007

DO - 10.1016/j.ijar.2015.07.007

M3 - Journal article

VL - 68

SP - 194

EP - 210

JO - International Journal of Approximate Reasoning

JF - International Journal of Approximate Reasoning

SN - 0888-613X

ER -