Improved Filter Feature Selection Methods by Samples Localization
With the development of high-throughput microarray technology, a large number of microarray data has been obtained by tens of thousands of simulation experiments on gene expression. However, due to the high cost, gene expression data of each experiment has a small number of samples for tens of thousands of genes. Furthermore, most of the genes, such as the features, are unrelated to classification. Therefore, developing an effective and robust method to extract informative genes subset from high dimensional microarray data is a challenging and important issue for microarray data analysis. In this article, we propose an improved feature selection method by samples localization. We only use localized samples to select genes subset, in order to effectively overcome the influence on the results of feature selection by outlier samples and the distribution of samples. Experimental results obtained by SVM classifiers on Leukemia dataset, Prostate dataset, Colon dataset, Breast dataset, DLBCL dataset and Lung dataset demonstrate that the proposed method is superior to those original filter feature selection methods both in best and average accuracies.
No Reference information available - sign in for access.
No Citation information available - sign in for access.
No Supplementary Data.
No Article Media
Document Type: Research Article
Publication date: August 1, 2013
More about this publication?
- Bionanoscience attempts to harness various functions of biological macromolecules and integrate them with engineering for technological applications. It is based on a bottom-up approach and encompasses structural biology, biomacromolecular engineering, material science, and engineering, extending the horizon of material science. The journal aims at publication of (i) Letters (ii) Reviews (3) Concepts (4) Rapid communications (5) Research papers (6) Book reviews (7) Conference announcements in the interface between chemistry, physics, biology, material science, and technology. The use of biological macromolecules as sensors, biomaterials, information storage devices, biomolecular arrays, molecular machines is significantly increasing. The traditional disciplines of chemistry, physics, and biology are overlapping and coalescing with nanoscale science and technology. Currently research in this area is scattered in different journals and this journal seeks to bring them under a single umbrella to ensure highest quality peer-reviewed research for rapid dissemination in areas that are in the forefront of science and technology which is witnessing phenomenal and accelerated growth.
- Editorial Board
- Information for Authors
- Subscribe to this Title
- Ingenta Connect is not responsible for the content or availability of external websites