To　predict　the　effect　of　heat　deviation　on　positioning　accuracy　of　feed　system　in　a　running　CNC　machine,　for　a　precision　coordinate　boring　machine,　the　temperature　accuracy　in　every　measuring　point　of　the　system　is　measured　by　infrared　thermal　imager　and　the　positioning　accuracy　is　measured　by　laser　interferometer　simultaneously,　then　the　least　squares　support　vector　machine　(LS-SVM)　prediction　method　of　thermal　error　in　feed　system　is　adopted　to　establish　a　prediction　model　of　temperature　and　location.　LS-SVM　prediction　method　is　introduced　into　the　prediction　model　to　predict　the　thermal　error　in　the　feed　system.　The　model　well　describes　the　nonlinear　relationship　between　the　thermal　errors　of　position　and　temperature　with　slight　dependence　on　samples　and　good　generalization　ability.　The　difficulty　that　the　prediction　error　gets　greater　because　the　average　characteristic　temperature　is　considered　instead　of　the　current　measuring　point　temperature　in　establishing　linear　prediction　model　is　solved.　The　results　show　that　the　prediction　precision　of　thermal　error　by　LS-SVM　model　reaches　90%,　30%　higher　than　that　by　the　existing　linear　prediction　model　for　NC　machine.　©　2015,　Editorial　Office　of　Journal　of　Xi＇an　Jiaotong　University.　All　right　reserved.