Knowledge discovery from granule features mining

Jianhong Luo,XiYong Zhu,Xiaojun Wang

DOI: https://doi.org/10.1007/978-3-642-38391-5_40

2013-01-01

Abstract:For the learning problem on imbalanced distribution of data sets, traditional machine learning algorithms tend to produce poor predictive accuracy over the minority class. In this paper, granule features mining model (GFMM) for knowledge discovery is proposed to improve classification accuracy on the minority class. Suitable information granules (IGs) are constructed by ETM-ART2, and then key features analysis method is proposed to discrete represent the IGs to mining compact knowledge rules. The final class for new samples inputted to GFMM can soon be decided by the knowledge rules. Experiments were conducted on three data sets with different skewed level, the results show that GFMM can lead to significant improvement on classification performance for the minority class and outperforms individual SVM and C4.5. © 2013 Springer-Verlag Berlin Heidelberg.