2013MNRAS.430..986J

With the commissioning year of the Large Sky Area Multi-Object Fibre Spectroscopic Telescope (LAMOST), the data archive of one-dimensional spectra is being released gradually. Searching for special objects like cataclysmic variables (CVs) in the data is one of LAMOST's objectives. This paper presents a novel method to identify CVs from optical spectra by using the support-vector machine (SVM) technique combined with principal-component analysis (PCA). After dimension reduction and feature extraction by PCA, spectral data are classified by SVM and most non-CVs are excluded. The final reduced list can be identified manually or by a template-matching algorithm. Experiments show that this data-mining method can find CVs from the LAMOST data base in an effective and efficient manner. We report the identification of 10 cataclysmic variables, of which two are new discoveries. In addition, this method is also applicable to mining other special celestial objects in sky-survey telescope data.