doi:

DOI: 10.3724/SP.J.1004.2008.00298

Acta Automatica Sinica (自动化学报) 2008/34:3 PP.298-304

Generalized Rough Set Method for Ensemble Feature Selection and Multiple Classifier Fusion


Abstract:
For improving the performance of multiple classifier system, a novel method of ensemble feature selection is proposed based on generalized rough set. In the paper, the relative dominance decision reduct (RDDR) with respect to multiple decision tables is presented to obtain the best feature subsets and interclass separability from different feature spaces. Then, the ensemble attribute reduction (EAR) method is proposed for ensemble feature selection. Using the KD-DWV algorithm based on knowledge discovery, the effectiveness of EAR was examined with the vegetation classification on a hyperspectral image. The result of the comparison experiment shows that EAR can be used to improve the generalization of multiple classifier system by combining appropriate multiple classifier fusion algorithm.

Key words:Ensemble feature selection, multiple classifier fusion, generalized rough set, hyperspectral

ReleaseDate:2014-07-21 14:25:59

Funds:Supported by National Natural Science Foundation of China (60574033), National Basic Research Program of China (973 Program)



1 Kittler J, Hatef M, Duin R P W, Matas J. On combining classifiers. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1998, 20(3): 226-239

2 Tsymbal A, Pechenizkiy M, Cunningham P. Diversity in search strategies for ensemble feature selection. Information Fusion, 2005, 6(1): 83-98

3 Kryszkiewicz M. Comparative study of alternative types of knowledge reduction in inconsistent systems. International Journal of Intelligent Systems, 2001, 16(1): 105-120

4 Swiniarski R W, Skowron A. Rough set methods in feature selection and recognition. Pattern Recognition Letters, 2003, 24(6): 833-849

5 Skowron A. Extracting laws from decision tables: a rough set approach. Computational Intelligence, 1995, 11(2): 371-388

6 Xiao Di, Hu Shou-Song. Real rough set theory and attribute reduction. Acta Automatica Sinica, 2007, 33(3): 253-258 (肖迪, 胡寿松. 实域粗糙集理论及属性约简. 自动化学报, 2007, 33(3): 253-258)

7 Zhang Wen-Xiu, Mi Ju-Sheng, Wu Wei-Zhi. Knowledge reductions in inconsistent information systems. Chinese Journal of Computers, 2003, 26(1): 12-18 (张文修, 米据生, 吴伟志. 不协调目标信息系统的知识约简. 计算机学报, 2003, 26(1): 12-18)

8 Sun Liang. Studies on Knowledge Discovery Methods for Fusion and Classification of High-dimensional Remote Sensing Data [Ph.D. dissertation], Xi'an Jiaotong University, 2006 (孙亮. 高维遥感数据融合与分类的知识发现方法研究 [博士学位论文], 西安交通大学, 2006)

9 Polkowski L. Rough Set: Mathematical Foundations. Heidelberg: Physica-Verlag, 2002. 23-24

10 Sun Liang, Han Chong-Zhao. Dynamic weighted voting for multiple classifier fusion: a generalized rough set method. Journal of System Engineering and Electronics Technology (English Edition), 2006, 17(3): 487-494

11 ENVI tutorial: vegetation hyperspectral analysisi [Online], available:http://www.rsinc.com/Envi/docs/tutorials/~Vegetation\_Analysis.pdf, May 30, 2007

12 Zhang Y, Mita D, Zhang J P, Jin M. Adaptive subspace decomposition for hyperspectral datadimensionality reduction. In: Proceedings of IEEE International Conference on Image Processing. Kobe, Japan: IEEE, 1999. 326-329

13 Richards J A, Jia X P. Remote Sensing Digital Image Analysis: An Introduction (Third Edition). Berlin, Germany: Springer-Verlag, 1999. 331

PDF