According　to　the　nine　main　factors,　which　affect　building　settlements　due　to　earthquake　liquefaction,　a　method　is　proposed　to　predict　building　settlements　due　to　earthquake　liquefaction　based　on　support　vector　regression　(v-SVR)　and　genetic　algorithm　(GA).　Since　the　modeling　of　this　method　uses　the　genetic　algorithm　to　automatically　determine　the　optimal　parameters　of　v-SVR,　and　it　is　directly　based　on　60　real　measured　seismic　settlement　samples,　the　nonlinear　relation　between　building　settlements　and　the　various　factors　is　established.　The　other　10　examples　are　predicated　by　the　training　model　to　achieve　good　results　with　the　average　relative　error　of　around　5%　compared　with　the　actual　building　settlements.　The　effectiveness　and　feasibility　have　been　proven.　The　analytic　method　and　process　discussed　in　this　paper　can　also　be　applied　to　the　seismic　damage　prediction　of　other　structures　of　different　forms.