This　paper　presents　least　squares　support　vector　machine　(LS-SVM)　neural　networks　as　a　new　tool　to　develop　the　model　of　the　Switched　Reluctance　Starter/Generator　(SRS/G).　A　LS-SVM　current-flux-linkage-position　model　and　a　LS-SVM　torque　model　are　then　used　to　simulate　the　dynamic　performance　of　SRS/G,　and　the　accuracy　of　the　model　is　tested　via　comparison　to　the　measurements　of　steady　state　waveforms.