Resumo

FERNANDEZ HERNANDEZ, Yumilka Bárbara et al. Effect of the features selection in the classification based on prototypes. Rev cuba cienc informat [online]. 2016, vol.10, n.4, pp.83-96. ISSN 2227-1899.

ABSTRACT Feature selection is a preprocessing technique with the objective of finding a subset of attributes that improves the classifier performance. In this paper is proposed a new method for solving classification problems based on prototypes (NP-BASIR-Class method) using feature selection. When using similarity relations for the granulation of the universe, similarity classes are generated, and a prototype is constructed for each similarity class. The feature selection method used was REDUCT-SIM based in the technique of optimization PSO (Particle Swarm Optimization). The main contribution of this investigation is demonstrating the utility of combining feature selection together to the prototype generation. The proposed algorithm was proven in groups of international data set and it was compared with well-known algorithms for the generation of prototypes. The experimental results show that the proposed method obtained satisfactory results, being the main advantage that is possible to reduce in the data set, the quantity of objects and the quantity of features obtaining satisfactory results without varying significantly the quality of the classification compared with the original data set.

Palavras-chave : features selection; prototype generation; similarity relations; classification.

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )