Oil-paper　insulation　is　the　key　insulation　structure　of　the　transformers,　whose　aging　condition　is　closely　related　to　the　operations　of　the　equipment.　The　degree　of　polymerization　(DP)　is　the　direct　parameter　characterizing　the　aging　condition　of　oil-paper　insulation.　Recently,　the　near　infrared　spectroscopy　(NIRS)　measurement　powered　by　quantitative　analysis　is　used　to　evaluate　DP　of　the　insulating　papers　mainly.　While　the　applications　of　NIRS　for　DP　evaluation　is　constrained　because　the　present　spectral　quantitative　analysis　method　is　not　accurate　and　stable　enough,　especially　for　onsite　tests.　In　this　paper,　we　propose　a　Gaussian　process　regression　(GPR)　method　to　predict　DP　of　the　oil-　paper　insulation　in　laboratory　as　well　as　in　field.　Firstly,　the　basic　principles　of　GPR　algorithm　are　illustrated.　A　GPR　model　for　DP　prediction　is　established　based　on　the　spectra　of　differently　aged　insulating　paper　samples　which　are　prepared　in　laboratory.　The　GPR　model　show　a　high　prediction　accuracy　both　for　training　set　and　testing　set.　The　established　GPR　model　is　finally　applied　on　the　DP　prediction　of　insulating　papers　originating　from　a　de-tanked　transformer.　The　accurate　predicted　DP　and　the　reliable　aging　assessment　results　indicate　that　the　established　model　can　be　implemented　on　site.　©　2021　IEEE.