@inproceedings{82972560518d4993a4770491bc8ff227,
title = "Noise-based feature perturbation as a selection method for microarray data",
abstract = "DNA microarrays can monitor the expression levels of thousands of genes simultaneously, providing the opportunity for the identification of genes that are differentially expressed across different conditions. Microarray datasets are generally limited to a small number of samples with a large number of gene expressions, therefore feature selection becomes a very important aspect of the microarray classification problem. In this paper, a new feature selection method, feature perturbation by adding noise, is proposed to improve the performance of classification. The experimental results on a benchmark colon cancer dataset indicate that the proposed method can result in more accurate class predictions using a smaller set of features when compared to the SVM-RFE feature selection method.",
keywords = "Classification, Feature perturbation, Gene selection, Microarray gene expression data",
author = "Li Chen and Goldgof, {Dmitry B.} and Hall, {Lawrence O.} and Eschrich, {Steven A.}",
year = "2007",
doi = "10.1007/978-3-540-72031-7_22",
language = "English (US)",
isbn = "3540720308",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "237--247",
booktitle = "Bioinformatics Research and Applications - Third International Symposium, ISBRA 2007, Proceedings",
note = "3rd International Symposium Bioinformatics Research and Applications, ISBRA 2007 ; Conference date: 07-05-2007 Through 10-05-2007",
}