Imbalanced Training Set Reduction and Feature Selection Through Genetic Optimization
R. Barandela, J. K. Hernández, J. Salvador Sánchez, Francesc J. Ferri · Conference on Artificial Intelligence Research and Development · 2005
Despite its simplicity and good classification performance, the Nearest Neighbor (NN) rule is not applied in many practical tasks because of the high amount of computational resources that it requires. Besides, when working with imbalanced training samples, its classification accuracy can be seriously degraded. In the present paper we propose two genetic algorithms to cope with these two issues. The purpose is to obtain complexity reduction while at the same time, to get a better balance in the training sample. Experimental results showing the benefits of our proposals are also reported.