Since 2008 August the Fermi Large Area Telescope (LAT) has provided a continuous coverage of the gamma-ray sky yielding more than 5000 γ-ray sources, but 54% of the detected sources remain unidentified or with no certain association with a low energy counterpart. Rigorous determination of class type for a γ-ray source requires the optical spectrum of the correct counterpart but optical observations are demanding and time-consuming, then machine learning techniques can be a powerful alternative for screening and ranking. We use machine learning techniques to select blazar candidates among uncertain sources characterized by γ-ray properties very similar to those of Active Galactic Nuclei. Consequently, the percentage of sources of uncertain type drops from 54% to less than 12% predicting a new zoo for the Fermi γ-ray sources. The result of this study opens up new considerations on the population of the gamma energy sky, and it will facilitate the planning of significant samples for rigorous analysis and multi-wavelength observational campaigns.

自2008年8月以来，费米大面积望远镜（LAT）连续覆盖了产生超过5000个γ射线源的伽马射线天空，但是54％的探测源仍未被识别或与低能对应物没有一定关联。严格确定γ射线源的类别类型需要正确对等物的光谱，但是光学观察是苛刻且耗时的，因此机器学习技术可以成为筛选和分级的有力替代方法。我们使用机器学习技术在以γ为特征的不确定来源中选择blazar候选对象射线特性与主动银河核非常相似。因此，不确定类型源的百分比从54％下降到不足12％，这预示着费米γ射线源的新动物园。这项研究的结果为伽马能天空的数量开辟了新的思路，并将有助于对重要样品的计划进行严格的分析和多波长观测活动。