Transductive Support Vector Machines for Cancer Diagnosis and Classification of Microarray Gene Expression Data

Robinson SemoliniFernando J. Von Zuben

The purpose of this paper is to present a powerful methodology for classification using gene expression data. The following problems will be considered: determination of specific diagnosis categories using the small round blue cell tumors of childhood, and classification of genes in functional groups. The classification task will be performed by means of Transductive Inference with SVM, that will be implemented based on training and testing data sets. In the case of training samples, experts have already previously classified the genes or diagnostic categories in their respective classes. So, given the testing samples, the purpose is to determine the corresponding class. Thereby, we will be able to classify in only one step each sample of the testing data set as a member or not of a certain class. As the transductive inference has not been previously applied to the classification of gene expression data, considering SVM as the classification tool, it will be compared with the traditional inductive method in a series of exhaustive experiments, with promising results.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: