.: Marcos Gestal :.     -----


Datos de la Publicación:

Carlos Fernandez-Lozano, José Antonio Seoane, Marcos Gestal, Tom R. Gaunt, Julián Dorado, Colin Campbell
Título: Texture classification using feature selection and kernel-based techniques
Revista: Soft Computing
ISSN: 1432-7643
Volumen: Vol 19(9)
Páginas: 2469-2480
Editorial: Springer-Verlang
Fecha Publicación: Septiembre 2015
Factor de Impacto: 1.304
SCIMago Journal Rank: 1.020
doi: http://dx.doi.org/10.1007/s00500-014-1573-5
Categorías WoS: Computer Science, Artificial Intelligence - Cuartil: Q2 - Tercil: - Posición 59 de 121
Computer Science, Interdisciplinary Applications - Cuartil: Q3 - Tercil: - Posición 59 de 102


The interpretation of the results in a classification problem can be enhanced, specially in image texture analysis problems, by feature selection techniques, knowing which features contribute more to the classification performance. This paper presents an evaluation of a number of feature selection techniques for classifcation in a biomedical image texture dataset (2-DE gel images), with the aim of studying their performance and the stability in the selection of the features. We analyze three different techniques: subgroup based Multiple Kernel Learning (MKL), which can perform a feature selection by down-weighting or eliminating subsets of features which shares similar characteristic, and two different conventional feature selection techniques such as Recursive Feature Elimination (RFE), with different classifiers (Naive Bayes, Support Vector Machines, Bagged Trees, Random Forest and Linear Discriminant Analysis), and a Genetic Algorithm-based approach with an SVM as decision function. The different classifiers were compared using a 10 times 10-fold cross validation model, and the best technique found is SVM-RFE, with an AUROC score of (95.88 ± 0.39%). However, this method is not significantly better than RFE-TREE, RFE-RF and grouped MKL, whilst MKL uses lower number of features, increasing the interpretability of the results. MKL selects always the same features, related with wavelet based textures, while RFE methods focuses specially co-occurrence matrix based features, but with high instability in the number of features selected.