Resumen de Using multidimensional projection to discover relations in attribute-based data
Eduardo Pérez Pérez, Larry A. Rendell
When shortage of domain knowledge prevents us from using good attributes to represent entities in a database, data mining and learning become especially difficult. If the only available attributes are primitive, their individual contribution to the target concept becomes insignificant and greedy learning methods fail. The the learner needs to discover relations among attributes. This paper purports the relational operator projection as a useful tool to find relations. A learning method based on multidimensional relational projection, MRP, is empirically compared with well-established machine learning methods. In spite of its simple search strategy, MRP performs comparatively well on synthetic concepts and real-world data. This advantage is attributed to the projection operator achieving the required functionality: finding relations.
