Applying the Support Vector Machine Method to Matching IRAS and SDSS Catalogues

This paper presents results of applying a machine learning technique, the Support Vector Machine (SVM), to the astronomical problem of matching the Infra-Red Astronomical Satellite (IRAS) and Sloan Digital Sky Survey (SDSS) object catalogues. In this study, the IRAS catalogue has much larger positional uncertainties than those of the SDSS. A model was constructed by applying the supervised learning algorithm (SVM) to a set of training data. Validation of the model shows a good identification performance (~ 90% correct), better than that derived from classical cross-matching algorithms, such as the likelihood-ratio method used in previous studies.