Ignitable liquid references (ILR), analysed by GC-MS by the Swedish National Forensic Centre (NFC), was subjected to OPLS-DA in order to investigate the possibility to classify ILR and further to use the generated models to predict analytical data from HAPSITE, a portable GC-MS instrument. The GC-MS raw data files obtained from NFC were processed through a library containing the most common compounds in IRL and the resulting dataset was modelled according to a hierarchic decision tree in order to separate different classes of ILRs. A selection of ILR was analysed by HAPSITE and the resulting data was used as a prediction-set to classify the IRLs by the generated OPLS-DA models. We were able to generate OPLS-DA models based on NFC's reference data that could classify different types of ILRs and to use those models to predict the identity of ILRs analysed by HAPSITE. These results show the potential to use models - generated on a large sample set such as NFC's reference data - on samples analysed by portable instruments. This could possibly be used to obtain faster identification of possible ILRs in the field.