According　to　the　complex　nonlinear　relationship　of　seismic　subsidence,　seven　measured　indicators　of　earthquake　liquefaction　were　selected　as　key　impacting　indicators　to　predict　the　building　settlements,　including　the　earthquake　intensity　I,　the　length-height　ratio　L/H,　the　average　contact　pressure　p,　the　width-depth　ratio　DB,　the　relative　density　RD,　the　thickness　of　non-liquefied　layer　and　the　depth　of　gravel　layer　wd.　Firstly,　41　groups　typical　cases　were　collected　as　the　training　sample,　and　20　groups　typical　cases　were　defined　as　the　test　sample,　which　has　enter　the　test　sample　and　output　the　prediction　accuracy　from　the　LS-SVM.　Secondly,　the　SVM　method　and　BP　method　were　chose　to　verify　the　validity　and　reliability　of　LS-SVM　model.　Finally,　a　example　analysis　show　that　the　method　is　feasible,　effective　and　simple　to　implement,　and　the　process　discussed　in　this　paper　can　also　be　applied　to　the　seismic　damage　prediction　of　other　structures　of　different　forms.　©　2015　ejge.