Incorporating Knowledge in Evolutionary Prototype Selection Salvador Garc´ ıa 1 , Jos´ eRam´onCano 2 and Francisco Herrera 1 1 University of Granada, Department of Computer Science and Artificial Intelligence, E.T.S.I. Inform´atica, 18071 Granada, Spain E-mail:{salvagl, herrera}@decsai.ugr.es 2 University of Ja´ en, Department of Computer Science, 23700 Linares, Ja´ en, Spain E-mail:jrcano@ujaen.es Abstract. Evolutionary algorithms has been recently used for proto- type selection showing good results. An important problem in prototype selection consist in increasing the size of data sets. This problem can be harmful in evolutionary algorithms by deteriorating the convergence and increasing the time complexity. In this paper, we offer a preliminary pro- posal to solve these drawbacks. We propose an evolutionary algorithm that incorporate knowledge about the prototype selection problem. This study includes a comparison between our proposal and other evolutionary and non-evolutionary prototype selection algorithms. The results show that incorporating knowledge improve the performance of evolutionary algorithms and considerably reduce time execution. 1 Introduction Most machine learning methods use all examples from the training data set. How- ever, data sets may contain noisy examples, that make the performance worse of these methods, or they may contain great amount of examples, increasing the complexity of computation. This fact is important especially for algorithms such as the k-nearest neighbors (k-NN) [1]. Nearest neighbor classification is one of the most well known classification methods in the literature. In its standard formulation, all training patterns are used as reference patterns for classifying new patterns. Instance selection (IS) is a data reduction process applied as preprocessing in data sets which are used as inputs for learning algorithms [2]. We consider data as stored in a flat file and described by terms called attributes or features. Each line in the file consists of attribute-values and forms an instance. By selecting instances, we reduce the number of rows in the data set. When we use the selected instances for direct classification with k-NN, then the IS process is called Prototype Selection (PS). Various approaches were proposed in order to carry out PS process in the lit- erature, see [3] and [4] for review. Evolutionary Algorithms (EAs) have been used to solve the PS problem with promising results [5][6]. These papers show that EAs outperform the non-evolutionary ones obtaining better instance reduction rates and higher classification accuracy. However, the increasing of the size of