Whereas　the　back　propagation　(BP)　neural　network　may　easily　go　to　the　local　minimum　value　in　the　optimization　process,　a　genetic　algorithm　based　BP　(GABP)　neural　network　was　constructed　by　combining　the　genetic　algorithm　(GA)　with　the　BP　neural　network.　The　characteristic　of　searching　the　group　optimization　for　GA　can　prevent　the　GABP　neural　network　from　converging　in　local　optimal　solution　and　ensure　that　it　finds　the　global　optimum　or　second-best　solution　with　good　performance.　The　training　of　the　GABP　neural　network　was　finished　in　two　steps.　The　GA　was　firstly　used　to　make　a　thorough　searching　in　the　global　space　for　the　weights　and　thresholds　of　the　neural　network,　which　can　ensure　they　fall　into　the　neighborhood　of　global　optimal　solution.　Then,　in　order　to　improve　the　convergence　precision,　the　gradient　method　was　used　to　finely　train　the　network　and　find　the　global　optimum　or　second-best　solution　with　good　performance.　The　experimental　data　of　an　orthogonal　design　(temperature,　pressure,　concentration)　for　a　micro-filtration　device　(1　μm　hydrophilic　polyvinylidene　fluoride　micro-filtration　membrane　for　bovine　serum　albumin　filtration)　were　used　as　the　sample　data　for　training　the　GABP　neural　network,　so　the　well-trained　network　can　be　used　to　predict　the　flux　of　the　micro-filtration　devices.　The　results　showed　that　this　method　had　greatly　improved　the　convergence　speed　and　the　prediction　accuracy　of　the　traditional　BP　network.