Velasco-Forero, Santiago A.

Publication Search Results

Now showing 1 - 1 of 1

Clasificación noparamétrica en datos direccionales
(2004) Velasco-Forero, Santiago A.; Acuña-Fernández, Edgar; College of Arts and Sciences - Sciences; Lorenzo, Edgardo; Vásquez, Pedro; Department of Mathematics; Segarra, Rafael
In a supervised classification problem, when the vectors of data are direction- al, it means, that they take values on a k-dimensional sphere, the application of the algorithms of pattern recognition as k-nearest-neighbour method, discriminant analysis, and kernel discriminant analysis, do not obtain good results in classification error rate. For this type of problems, we propose several algorithms based on directional k-nearest-neighbour, estimation of density for directional kernel and discriminant analysis with assumption of von Mises-Fisher distribution. Additionally we present an extension for these classification methods for directional data in standard sets (not directional), based in the correlation matrix. We illustrate the performance of these methods on simulated data, machine learning datasets and microarray data sets.