The　effective　prediction　of　housing　prices　is　of　great　significance　to　buyers　and　real　estate　developers.　However,　most　existing　traditional　approaches　are　inefficient　and　low　accuracy.　In　this　paper,　we　forecast　the　growth　rate　of　house　prices　in　different　levels　of　cities　of　China　by　using　Dynamic　Model　Averaging　and　Dynamic　Model　Selection,　which　allows　both　parameters　as　well　as　forecasting　model　rapid　change.　In　addition,　some　evidence　show　that　Internet　search　indexes　could　help　model　to　improve　prediction　accuracy,　so　we　have　introduce　Internet　search　indexes　by　using　some　key　words　under　the　background　of　big　data　in　recent　years.　Under　the　cirteria　of　MAFE,　MSFE　and　MLPL,　the　results　of　model　forecast　performance　indicate　that　DMA　and　DMS　model　achieve　significantly　forecasting　accuracy　especificlly　forgetting　factors　alpha　=　lambda　=　0.95.