Using a clustering similarity measure for feature selection in high dimensional data sets

Feature selection is a very important preprocessing step in data classification. By applying it we are able to reduce the dimensionality of the problem by removing redundant or irrelevant data. High dimensional data sets are becoming usual nowadays specially in bio-informatics, biology, signal processing or text classification, increasing the need for efficient feature selection methods. In this paper we study the applicability of a clustering validation measure, the Adjusted Rand Index (ARI), for this task comparing it with other methods based on statistical tests and on ROC curve. We have performed some experiments that show the validity of the proposed method.