Supervised Automatic Identification of Extragalactic Sources in the WISExSUPERCOSMOS Catalogue

Abstract

We present new catalogue of ∼8,500,000 extragalactic objects as a result of automatic classification of WISE and SuperCOSMOS (SCOS) cross-identification product. The main goal is to create a set of candidates in extragalactic objects due to colour (photometric) features through machine learning techniques. Extragalactic sources were separated from stars in high-dimensional colour space using Support Vector Machine (SVM) classifier. Construction of catalogue of the extragalactic objects is based on the four important procedures: 1. Cross-identification of the WISExSCOS catalogues. 2. Training set creation (Gaia DR1 and 2MASX/XSC data). 3. Feature engineering and colour-space constructing for further learning and classification. 4. Fine-tuning of SVM and separation and classification processes. In result we got high-accuracy (∼98%) algorithm for extragalactic source identification in built colour space. Product of algorithm realization is presented as photometric catalogue of the extragalactic objects and can be used for further astronomical investigations.