An　accurate　tumor　classification　is　important　to　diagnosis　and　treatment　cancers.　The　conventional　methods　for　tumor　classification　include　training　and　testing　phases,　which　may　cause　over　fitting.　Although　this　problem　can　be　avoided　by　using　sparse　representation　classification,　the　existing　sparse　representation　methods　for　tumor　classification　are　inefficient.　In　this　paper,　an　efficient　and　robust　classification　model　LSRC　based　on　least　square　regression　and　nearest　subspace　rule　is　adopted　for　tumor　classification.　To　investigate　its　performance,　our　proposed　model　LSRC　is　compared　with　3　existing　methods　on　9　tumor　datasets.　The　experimental　results　show　that　our　proposed　model　can　use　less　time　to　achieve　higher　classification　accuracy.