In　this　paper,　a　new　nonlinear　classification　algorithm　is　proposed　using　Isomap　(ISOmetric　feature　MAPping).　First,　nonlinear　dimensionality　reduction　algorithm　is　implemented　using　Isomap;　then　nonlinear　mapping　matrix　is　approximated　by　local　linear　mapping;　next　dimensionality　reduction　for　all　data　sets　is　implemented　by　local　linear　mapping;　finally　linear　classification　is　implemented.　Experimental　results　confirm　the　efficiency　of　proposed　algorithm.