Author(s): , ,
Institution(s): 1. National Astronomical Observatories, 2. Wuhan University
Previous studies have shown that information from more bands by cross-matching different surveys is helpful to improve quasars' photometric redshift estimation. However the number of sample by cross-matching decreases and those
from single survey may not be estimated accurately. In this paper, we propose a new two-stage collaborative approach, including analysis stage and integration stage, to improve photometric redshift estimation accuracy. In the analysis stage, we apply decision-tree method, a data-driven algorithm, to systematically analyze conditions which derive accurate or inaccurate phototometric redshift estimation with the samples of SDSS and UKIDSS, respectively. Since the design principles of the two surveys are different. Consequently, the two instruments show accurate or inaccurate estimations in different conditions for different groups of celestial objects. In the integration stage, we combine these advantage conditions from both surveys together to form a collaborative plan. Experimental results show that the collaborative approach improves quasars' photometric redshift eastimation accuracy while covering more quasars than cross-matching approach.