Abstract
Conference Title: 2014 6th International Conference of Soft Computing and Pattern Recognition (SoCPaR) Conference Start Date: 2014, Aug. 11 Conference End Date: 2014, Aug. 14 Conference Location: Tunis, Tunisia For many years the dependencies between variables have constituted a challenge of many research area especially in data analysis and data mining. Since the datasets used in the studies are more complex and rich, the variables, that describe these data, become more structured and are often interlinked. When the interlinkage of variables, called " dependencies between variables", are taken into consideration and processed correctly by the used algorithms, this will generally eliminate incoherencies and improve the quality of the study results. In this paper we will explain how we treat the hierarchical variable dependencies in feature selection. The data used by our algorithm are represented by a set of Boolean Symbolic Objects (BSOs). A BSO is multi-valued object which can represent not only an individual, but a cluster of individuals.