Clustering technology has been used extensively for the analysis of gene expression data. Among various clustering methods, soft subspace clustering algorithms developed in recent years have demonstrated more promising performance than most traditional clustering algorithms and hard subspace clustering algorithms. Many soft subspace clustering algorithms have effectively utilized the within-cluster information, such as the within-cluster compactness, to develop the corresponding algorithms but few of them pay enough attention to other important information, such as the between-cluster information. Thus, it deserves further study to enhance soft subspace clustering by integrating more useful information in the clustering procedure. In this study, enhanced subspace clustering techniques are investigated for the clustering analysis of high dimensional gene expression data by integrating the within-cluster and between-cluster information simultaneously. First, a new optimization objective function is presented by integrating the fuzzy within-class compactness and the between-cluster separation in the weighting subspace. The corresponding learning rules for clustering are then derived based on the proposed objective function and a new soft subspace clustering algorithm, named as Enhanced Entropy-Weighting Subspace Clustering (EEW-SC), is proposed. The performance of the proposed algorithm on the clustering analysis of various high dimensional gene expression datasets is experimentally compared with that of several competitive subspace clustering algorithms. Our experimental studies demonstrate that the proposed algorithm can obtain better performance than most of the existing soft subspace clustering algorithms.
- ε-Insensitive robustness distance
- Clustering ensemble
- Entropy weighting
- Gene expression clustering analysis
- Subspace clustering
ASJC Scopus subject areas