Abstract:Unsupervised feature selection has been always attracting research attention in the communities of machine learning and data mining for decades. In this paper, we propose an unsupervised feature selection method seeking a feature coefficient matrix to select the most distinctive features. Specifically, our proposed algorithm integrates the Maximum Margin Criterion with a sparsity-based model into a joint framework, where the class margin and feature correlation are taken into account at the same time. To maximize the total data separability while preserving minimized within-class scatter simultaneously, we propose to embed Kmeans into the framework generating pseudo class label information in a scenario of unsupervised feature selection. Meanwhile, a sparsity-based model, ` 2 ,p-norm, is imposed to the regularization term to effectively discover the sparse structures of the feature coefficient matrix. In this way, noisy and irrelevant features are removed by ruling out those features whose corresponding coefficients are zeros. To alleviate the local optimum problem that is caused by random initializations of K-means, a convergence guaranteed algorithm with an updating strategy for the clustering indicator matrix, is proposed to iteractively chase the optimal solution. Performance evaluation is extensively conducted over six benchmark data sets. From plenty of experimental results, it is demonstrated that our method has superior performance against all other compared approaches.

Feature Selection Based on Classes Margin

Feature Selection Based on Dependency Margin

An Unsupervised Feature Selection Method Based on Improved ReliefF and Bisecting K-means

Feature Selection Method on Imbalanced Text

A Constrained Feature Selection Approach Based on Feature Clustering and Hypothesis Margin Maximization

Large-margin Feature Selection for Monotonic Classification

Unsupervised Feature Analysis with Class Margin Optimization

A Multi-Class Large Margin Classifier

Margin-maximizing feature elimination methods for linear and nonlinear kernel-based discriminant functions

A Survey for Study of Feature Selection Algorithms

An Adaptive Feature Selection Method for Multi-Class Classification.

Hyperspectral Data Classification Using Margin Infused Relaxed Algorithm

Feature Selection for Support Vector Machines in Text Categorization

Support Vector Machine-Recursive Feature Elimination for Localized Feature Selection

A Hybrid Feature Selection Approach by Correlation-Based Filters and SVM-RFE

Multiclass feature selection algorithms base on R-SVM

Multiclass Cancer Classification by Using Fuzzy Support Vector Machine and Binary Decision Tree with Gene Selection

Feature Selection and Classification of Polarimetric SAR Images Using SVM

Feature selection based on run covering

An Approximate Markov Blanket Feature Selection Algorithm

A feature-selection algorithm based on Support Vector Machine-Multiclass for hyperspectral visible spectral analysis