Dimensionality Reduction Approach using Attributes Extraction and Attributes Selection in Gene Expression Databases

The gene expression databases are formed by a high number of attributes. To deal with this amount, data dimensionality reduction is used in order to minimize the volume of data to be treated regarding the number of attributes, and to increase the generalization capability of learning methods by eliminating irrelevant and/or redundant data. This paper proposes an approach to means of dimensionality reduction, which joins attribute extraction and attributes selection. For this, we used the Random Projection method and the filter and wrapper approaches for the attribute selection. The experiments are realized in five gene expression microarray databases. The results of the experiments showed that join of those approaches can provide promising results.

Projetos relacionados