Locally linear embedding and neighborhood rough set-based gene selection for gene expression data classification

Abstract

ABSTRACT. Cancer subtype recognition and feature selection are important problems in the diagnosis and treatment of tumors. Here, we propose a novel gene selection approach applied to gene expression data classification. First, two classical feature reduction methods including locally linear embedding (LLE) and rough set (RS) are summarized. The advantages and disadvantages of these algorithms were analyzed and an optimized model for tumor gene selection was developed based on LLE and neighborhood RS (NRS). Bhattacharyya distance was introduced to delete irrelevant genes, pair-wise redundant analysis was performed to remove strongly correlated genes, and the wavelet soft threshold was determined to eliminate noise in the gene datasets. Next, prior optimized search processing was carried out. A new approach combining dimension reduction of LLE and feature reduction of NRS (LLE-NRS) was developed for selecting gene subsets, and then an open source software Weka was applied to distinguish different tumor types and verify the cross-validation classification accuracy of our proposed method. The experimental results demonstrated that the classification performance of the proposed LLE-NRS for selecting gene subset outperforms those of other related models in terms of accuracy, and our proposed approach is feasible and effective in the field of highdimensional tumor classification.

An Attribute Reduction Method Using Neighborhood Entropy Measures in Neighborhood Rough Sets

“…To further verify the classification performance of our proposed method, the eight methods were employed to evaluate the number of selected genes and the classification accuracy on the four gene expression data sets selected from Table 2 . The ARDNE algorithm was compared with the seven related state-of-the-art dimensionality reduction methods, which included: (1) The sequential forward selection algorithm (SFS) [ 51 ], (2) the sparse group lasso algorithm (SGL) [ 52 ], (3) the adaptive sparse group lasso based on conditional mutual information algorithm (ASGL-CMI) [ 53 ], (4) the Spearman’s rank correlation coefficient algorithm (SC 2 ) [ 44 ], (5) the gene selection algorithm based on fisher linear discriminant and neighborhood rough set (FLD-NRS) [ 39 ], (6) the gene selection algorithm based on locally linear embedding and neighborhood rough set algorithm (LLE-NRS) [ 40 ], and (7) the RelieF algorithm [ 41 ] combined with the NRS algorithm [ 49 ] (RelieF+NRS). The SVM classifier in the WEKA tool was used to do some simulation experiments.…”

“…Therefore, the time complexity of ARDNE is close to O ( mn ). So far, ARDNE appears to be more efficient than some of the existing algorithms for attribute reduction in [ 33 , 39 , 40 , 41 ] in neighborhood decision systems. Furthermore, its space complexity is O ( mn ).…”

Dimensionality reduction techniques for high dimensional data: State of the art

“…In recent years, gene selection and classification methods have been developed. A technique that combines LLE and Neighborhood Rough Set (NRS) was proposed by Sun et al [31]. The irrelevant genes were dropped by measuring the separability between each class.…”

Feature selection based on multiview entropy measures in multiperspective rough set

“…The results of MIFS and other 11 dimensionality reduction methods on these four data sets are compared. The 11 methods used for comparison are: the sequential forward selection algorithm (SFS), 49 the Spearman's rank correlation coefficient algorithm (SC2), 50 the sparse group Lasso algorithm (SGL), 51 the mutual information maximization algorithm (MIM), 49 the adaptive sparse group Lasso algorithm based on conditional mutual information (ASGL‐CMI), 52 the distributed ranking filter approach removing the features with information gain zero from the ranking and correlation‐based feature selection algorithm (DRFO‐CFS), 52 the neighborhood rough set‐based reduction algorithm (NRS), 32 locally linear embedding and neighborhood rough set (LLE‐NRS), 53 the Relief algorithm combined with the NRS algorithm (Relief‐NRS), 54 the gene selection algorithm based on Fisher linear discriminant and neighborhood rough set (FLD‐NRS), 18 and the fuzzy backward feature elimination algorithm (FBFE) 55 12 and 13 show, respectively, the size and SVM accuracy of the reduction subset selected by the 12 methods on the four data sets, where the meaning of the symbol “/” is the same as that in Tables 9–11.…”