In this paper we present a GP-based method forautomatically evolve projections, so that data can bemore easily classified in the projected spaces. At thesame time, our approach can reduce dimensionality byconstructing more relevant attributes. Fitness of eachprojection measures how easy is to classify the datasetafter applying the projection. This is quickly computedby a Simple Linear Perceptron. We have tested ourapproach in three domains. The experiments show that itobtains good results, compared to other MachineLearning approaches, while reducing dimensionality inmany cases
CITATION STYLE
Estebanez, C., Aler, R., & Valls, J. M. (2005). Genetic Programming Based Data Projections for Classification Tasks. In C. Ardil (Ed.), International Enformatika Conference, IEC’05 (Vol. 7, pp. 56–61). Prague, Czech Republic: Enformatika, Çanakkale, Turkey. Retrieved from http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.193.6069.pdf
Mendeley helps you to discover research relevant for your work.