This　paper　proposes　a　new　geometric　error　compensation　algorithm　for　the　five-axis　machine　tools.　Firstly,　the　general　machining　error　model　of　five-axis　machine　tool　is　established　based　on　multi-body　theory　(MBS)　and　homogeneous　transmission　matrix　(HTM).　Secondly,　considering　deviation　of　tool　setting　point,　the　initial　NC　code　of　each　axis　is　calculated.　Thirdly,　based　on　the　initial　NC　code,　a　geometric　error　compensation　is　developed　by　the　present　improved　particle　swarm　optimization　(IPSO)　to　seek　the　optimal　NC　code.　In　this　model,　the　tool　pose　and　the　volume　error　are　regarded　as　particles　and　optimal　object,　respectively.　According　to　the　objective　function,　the　optimal　particles　are　selected　in　a　given　interval　to　reduce　the　volume　error　of　machine　tool.　Finally,　an　experiment　is　implemented　by　the　S-shaped　test　to　verify　the　effectiveness　of　this　method.　Compared　with　linear　interpolation　(LI)　compensation,　the　comprehensive　machining　accuracy　based　on　PSO　method　has　improved　by　17.5%,　and　the　maximum　machining　error　has　reduced　by　0.0058　mm.