The　susceptibility　of　landslide　is　a　regional　landslide　spatial　prediction　and　is　the　natural　attribute　of　landslide　disasters.　It　closely　relates　to　every　factor　of　disaster-pregnant　environment.　Based　on　the　Elman　neural　network　algorithm,　a　trial　is　made　to　combine　it　with　gray　system　analysis　method.　Main　influencing　factors　of　landslides　were　objectively　selected　as　input　variables　of　Elman　neural　network,　and　the　optimization　problem　of　variable　selection　of　Elman　neural　network　was　resolved.　Two　landslide　sites　in　different　study　areas　(　Minging/　Minhou　counties　and　Jianning　County,　Fujian　Province)　were　trained　and　fitted　with　the　Elman　neural　network,　and　then　landslide　susceptibility　of　the　study　areas　was　forecasted　using　the　fitted　model.　A　good　prediction　result　was　obtained　using　the　proposed　methods.