Earthquake　is　one　visitation　of　providence　which　threatens　people＇s　safety.　Once　earthquakes　happen　in　a　city　there　will　be　great　losses.　How　to　find　a　city＇s　weak　link　so　as　to　provide　correlative　measure　to　mitigate　the　losses　is　necessary.　The　prediction　of　seismic　disaster　is　one　basic　work　of　that.　In　this　paper,　a　method　for　predicting　a　city＇s　buildings＇　seismic　disaster　is　presented.　The　general　survey　data　of　buildings　was　used　as　factors　affecting　buildings＇　seismic　disaster　and　an　artificial　neural　network　was　used　as　a　tool　to　calculate　the　seismic　disaster.　An　artificial　neural　network　model　was　made　because　this　type　of　model　has　an　ability　to　simulate　nonlinear　consequences.　Typical　destroyed　buildings　were　collected　as　samples　from　many　previous　earthquakes.　After　training　the　model　using　the　collected　data　a　convergent　model　was　acquired.　A　conclusion　was　drawn　that　the　model　was　sufficient　for　predicting　buildings＇　seismic　disaster　after　using　it　to　predict　the　seismic　disaster　of　a　group　of　new　buildings.