DOI: 10.7763/IJCTE.2011.V3.331
Extension of Learning Vector Quantization to Cost-sensitive Learning
Abstract—Learning vector quantization (LVQ) is an effective network model to solve classification tasks in a wide variety of real world applications. The usage of LVQ has been extended to hybrid data type. In this paper, we propose a weighted version of BNCLVQ, which incorporates the cost matrix into prototype learning and labeling by means of instance weighting. Empirical results show the superiority of proposed algorithm over original NBCLVQ and some variants on both binary-class data and multi-class data.
Index Terms—Classification, Cost-sensitive learning, Learning vector quantization, Hybrid data type.
Ning Chen is with GECAD, Instituto Superior de Engenharia do Porto, Instituto Politecnico do Porto (telphone: 351-22-8340500; fax: 351-22-8321159; email: ningchen74@gmail.com).
Bernardete Ribeiro is with CISUC, Department of Informatics Engineering, University of Coimbra, Portugal (email: bribeiro@dei.uc.pt).
Armando Vieira is with Instituto Superior de Engenharia do Porto, Instituto Politecnico do Porto (email: asv@isep.ipp.pt).
João Duarte is with GECAD, Instituto Superior de Engenharia do Porto, Instituto Politecnico do Porto (email: jmmd@isep.ipp.pt).
João C. Neves is with ISEG, School of Economics, Technical University of Lisbon, Portugal (email: jcneves@iseg.utl.pt).
Cite: Ning Chen, Bernardete Ribeiro, Armando Vieira, João Duarte, and João C. Neves, "Extension of Learning Vector Quantization to Cost-sensitive Learning," International Journal of Computer Theory and Engineering vol. 3, no. 3, pp. 352-359, 2011.