Thermal　error　compensation　of　high-speed　motorized　spindle　is　one　of　the　effective　methods　to　improve　the　machining　accuracy　of　CNC　machine　tools.　In　this　paper,　the　variation　law　of　the　thermal　error　of　the　motorized　spindle　is　studied,　a　method　for　selecting　temperature　sensitive　points　is　proposed,　and　the　thermal　error　prediction　model　is　established.　First,　the　temperature　and　thermal　error　of　the　motorized　spindle　at　different　speeds　are　measured　by　thermal　error　experiment.　Secondly,　the　clustering　by　fast　search　and　find　of　density　peaks　(CFSFDP)　is　used　to　solve　the　problem　that　the　optimal　cluster　number　is　unknown　in　the　process　of　selecting　temperature　sensitive　points,　and　the　temperature　sensitive　points　under　multi　speed　of　the　motorized　spindle　are　obtained.　Finally,　the　adaptive　boundary　harris　hawk　algorithm　(ABHHO)　is　used　to　optimize　the　hyperparameters　of　the　least　squares　support　vector　machine　(LSSVM),　and　a　prediction　model　for　the　thermal　error　based　on　ABHHO-LSSVM　is　established,　which　improves　the　prediction　accuracy.　The　proposed　method　and　model　provide　a　technical　basis　for　thermal　error　compensation　of　motorized　spindle.