Traditional　kernelised　classification　methods　could　not　perform　well　sometimes　because　of　the　using　of　a　single　and　fixed　kernel,　especially　on　some　complicated　data　sets.　In　this　paper,　a　novel　optimal　double-kernel　combination　(ODKC)　method　is　proposed　for　complicated　classification　tasks.　Firstly,　data　sets　are　mapped　by　two　basic　kernels　into　different　feature　spaces　respectively,　and　then　three　kinds　of　optimal　composite　kernels　are　constructed　by　integrating　information　of　the　two　feature　spaces.　Comparative　experiments　demonstrate　the　effectiveness　of　our　methods.　©　2009　Springer　Berlin　Heidelberg.