Nonuniform　indoor　temperature　distributions　have　been　widely　introduced　due　to　the　increasing　concerns　brought　about　by　the　trade-off　between　building　energy　consumption　and　thermal　comfort.　Related　to　this,　obtaining　real-time　indoor　temperature　field　is　crucial　for　control　and　regulation.　Previously,　a　rapid　temperature　prediction　algorithm　based　on　the　contribution　ratio　of　indoor　climate　(CRI)　and　sensor　information　was　pro-posed.　Compared　with　fixed　sensors,　mobile　sensors　have　better　moving　flexibilities　and　higher　prediction　ac-curacies.　Given　that　considerable　data　can　be　obtained　from　one　sensor,　the　chosen　mobile　sensor　location　has　a　strong　correlation　with　the　prediction　accuracy.　However,　only　a　few　studies　have　investigated　the　impact　of　the　sensor　location.　Thus,　this　study　aimed　to　determine　the　optimal　mobile　sensor　location　for　improving　prediction　accuracy　using　the　proposed　algorithm.　Two　differently　scaled　rooms　were　numerically　built,　and　1.2　m　high　points　were　set　as　the　selected　locations.　After　experimental　validation　of　the　simulation　results,　three　factors　(moving　path,　supply/return　air,　and　air　velocity)　were　investigated　to　determine　their　effects　on　prediction　accuracy.　The　results　showed　that　the　air　velocity　was　the　only　significantly　impactful　factor.　Furthermore,　two　principles　of　choosing　location　were　proposed:　(1)　when　the　percentage　of　dominant　velocity　is　high,　all　points　should　have　one　identical　dominant　direction,　and　(2)　when　the　percentage　of　dominant　velocity　is　low,　points　should　have　a　similar　actual　direction.　These　principles　were　verified　in　both　models　with　a　prediction　accuracy　greater　than　85%.