Thermal　error　is　the　key　error　source　which　affects　the　stability　of　machining　accuracy　and　temperature　measurement　point　optimization　is　the　premise　and　difficult　to　achieve　accurate　thermal　error　compensation.　In　order　to　reduce　the　temperature　measurement　point　and　improve　the　efficiency　of　temperature　data　acquisition,　The　method　is　proposed　based　on　principal　component　analysis　(PCA)　and　Single　Variable　Contribution　(SVC).　This　method　is　mainly　to　reduce　the　temperature　measurement　points　of　the　machine　tool,　so　as　to　facilitate　the　later　thermal　error　modelling.　firstly,　PCA　is　introduced　to　extract　the　principal　components　of　temperature　data　samples.　Then,　the　contribution　of　each　temperature　variable　attribute　in　the　main　component　space　is　recorded　as　a　single　variable　contribution　degree,　and　based　on　which　the　key　measuring　points　is　screened　out.　Thirdly,　multivariable　linear　regression　model　is　used　as　the　thermal　error　model　of　the　spindle　in　z-axix.　Finally,　three　axis　CNC　machine　tool　mu　1000/3v　is　taken　as　the　research　object　and　the　results　show　that　the　proposed　method　can　reduce　the　temperature　measuring　point　from　24　to　6,　and　minimum　residual　error　is　only　0.5　mu　m.　This　fully　shows　that　the　method　proposed　in　this　paper　is　effective,　and　has　a　certain　guiding　role　to　improve　the　precision　prediction　of　machine　tool.