A Method Based on Genetic Programming for Improving the Quality of Datasets in Classification Problems

摘要: The problem of the representation data is a key issue in Machine Learning (ML) field. ML tries to automatically induct knowledge from set examples or instances problem, learning how distinguish between different classes. It known that inappropriate representations can drastically limit performance algorithms. On other hand, high-quality same data, produce strong improvement classification rates. In this work we present GP-based method for evolve projections. These projections change space into higher-quality one, thus improving At time, our approach reduce dimensionality by constructing more relevant attributes. We have tested four domains. experiments show it obtains good results, compared approaches do not use projections, while reducing many cases.