In　order　to　optimize　the　operating　conditions　in　dead-end　microfiltration,　a　predicting　model　relating　the　specific　resistance　to　the　operating　conditions　based　on　the　BP　neural　network　was　firstly　developed　in　this　paper,　in　which　the　experimental　data　of　an　orthonormal　design　table　(53)　were　used　as　the　input　sample　data.　Then,　by　using　the　average　relative　absolute　error　of　the　testing　sample　as　the　criterion,　a　comparison　experiment　of　the　predicting　for　specific　resistance　of　yeast　suspensions　by　using　the　BP　neural　network　method　and　the　multiple　linear　regression　method　had　been　made.　Finally,　the　predicting　precisions　of　two　models　had　been　given.　The　results　showed　that　the　BP　neural　network　method　was　better　than　the　multiple　linear　regression　method　and　the　average　relative　absolute　errors　were　3.55%　and　5.16%　for　the　BP　neural　network　method　and　the　multiple　linear　regression　method,　respectively.