Abstract:

The small sample data in the high-dimensional data space are encountered in biological applications such as in gene expression microarrays and proteomics mass spectrometry. Due to the fact that such data have characteristics such as high-dimensionality and small sample dimension, their classification becomes hard. Many feature selection algorithms were developed for the purpose of reducing the dimensionality of this kind of data and improving the accuracy of classifiers. In the realization of area discoveries through feature sets, the selected feature subsets skip important information in unnecessary feature sets. This problem comes into prominence with the feature, in the process of performing the discovery of information from the high-dimensional data space. This paper evaluates the proposed ensemble gene selection method based on a local feature selection to Artificial Immune Recognition algorithms in order to find the optimal biological sequences. The unique feature of this study is developing the different type of associated feature groups defined using high-dimensional data in order to find important tumor-related genes. The comparative tests were performed on the training set and test set separately with using support vector machines and k-NN classifiers.

Keywords: