Data Set Editing by Ordered Projection

1 Department of Computer Science. University of Seville. Spain. {aguilar, riquelme, mtoro}@lsi.us.es Abstract. In this paper, an editing algorithm based on the projection of the examples in each dimension is presented. The algorithm, that we have called EOP (Editing by Ordered Projection) has some interesting characteristics: important reduction of the number of examples from the database; lower computational cost in respect of other typical algorithms due to the absence of distance calculations; conservation of the decision boundaries, especially from the point of view of the application of axis-parallel classifiers; reduction of the decision tree size or the number of decision rules. The performance of EOP is showed by comparing the results provided by C4.5 [5] before and after applying it on databases with continuous attributes. These experiments have been realised using some databases from UCI repository [1]. The use of EOP as preprocessing method for the later application of any axis-parallel learning algorithm convert it in a valuable tool in the field of data mining.