We present new diagnostic tools for distinguishing supernova remnants (SNRs) from H II regions. Up to now, sources with flux ratio [S II]/H{α } higher than 0.4 have been considered as SNRs. Here, we present combinations of three or two line ratios as more effective tools for the separation of these two kinds of nebulae, depicting them as 3D surfaces or 2D lines. The diagnostics are based on photoionization and shock- excitation models (MAPPINGS III) analysed with support vector machine (SVM) models for classification. The line-ratio combination that gives the most efficient diagnostic is [O I]/H{α } - [O II]/H{β } - [O III]/H{β }. This method gives 98.95 per cent completeness in the SNR selection and 1.20 per cent contamination. We also define the [O I]/H{α } SNR selection criterion and measure its efficiency in comparison with other selection criteria.
