The　accuracy　retainability　is　becoming　an　important　performance　index　of　machine　tool,　and　how　to　improve　it　is　a　tough　problem　faced　to　manufacturers　and　users.　Generally,　it　needs　to　measure　the　errors　termly　and　repeatedly　during　the　specified　period　to　analyze　the　timeliness　machining　accuracy　retainability,　which　generates　intricate　and　vast　error　data.　In　this　paper,　a　solution　to　predict　machining　accuracy　retainability　is　proposed　based　on　least　square　support　vector　machine　(LS-SVM).　A　vertical　machining　center　that　machines　plane　and　hole　continuously　for　half　a　year　is　selected　as　an　illustrative　example.　The　analysis　results　show　that　the　proposed　method　is　good　at　predicting　the　timeliness　machining　accuracy　retainability　of　machine　tool.