Estimating photometric redshifts of quasars using support vector machines

With various algorithms successfully applied for measuring photometric redshifts of galaxies, we utilize support vector machines (SVMs), an empirical training-set method, to estimate photometric redshifts of quasars by means of five-band photometry data from Sloan Digital Sky Survey (SDSS). Using a sample of 67,491 quasars from SDSS Data Release Five (SDSS DR5), we explore the influence of model parameters of SVMs on the accuracy of photometric redshifts of quasars. SVMs are trained on two thirds of sample, and tested on the rest sample. The variance between the photometric and spectroscopic redshifts is 0.119, and 48.94%, 70.71% and 78.12% of the objects are within Delta z<0.1, 0.2 and 0.3, respectively. Compared to the color-redshift-relation (CZR), SVMs show their superiority.