Abstract
Many clustering techniques have been proposed for the analysis of gene expression data. However, the optimal method for a given experimental dataset is still not resolved. Fuzzy c-means and kernel fuzzy c-means algorithm have been widely applied to gene expression data, but they give the equal weight to the genes and noises, which lead to results that are not stable or accurate. In this paper, we propose a local weighted fuzzy clustering method in the kernel space. The original data is mapped to the high-dimensional feature space and Gaussian function is employed to investigate the local information of the cluster centre. Consequently, it will assign different weights to the noise and genes. Our experiments show that the proposed methods achieve better clustering effect than the fuzzy clustering algorithm and fuzzy kernel clustering algorithm.
| Original language | English |
|---|---|
| Title of host publication | Engineering and Technology (S-CET), 2012 Spring Congress on |
| Place of Publication | Piscataway |
| Publisher | IEEE |
| ISBN (Electronic) | 978-1-4577-1964-6 |
| ISBN (Print) | 978-1-4577-1965-3 |
| Publication status | Published - 2012 |
Keywords
- clustering
- gene expression data
- kernel function
- noise