In　this　study,　the　support　vector　machine　(SVM)　model　which　was　based　on　restricted　data　sets　(the　size　of　the　training　set　is　small　or　small　training　sample)　was　applied　to　predict　the　permeate　flux　and　rejection　of　Bovine　serum　albumin　(BSA)　of　homemade　VC-co-VAc-OH　microfiltration　membrane　as　the　function　of　fabrication　conditions.　The　membrane　preparation　conditions　(the　solid　content,　the　additive　content,　environmental　temperature,　the　relative　humidity,　evaporation　time　of　a　volatile　solvent,　precipitation　temperature,　and　precipitation　time)　were　input　variables;　pure　water　flux　and　rejection　of　BSA　were　output　variables.　The　results　showed　that　the　detailed　relationships　between　fabrication　conditions　and　filtration　performance　of　the　membranes　could　be　established.　Excellent　agreements　between　the　prediction　of　SVM　model　and　the　experiments　validate　that　SVM　model　has　sufficient　accuracy.　Furthermore,　the　results　predicted　by　SVM　model　were　compared　with　those　predicted　by　artificial　neural　network　(ANN)　model　which　was　widely　used　in　the　optimization　of　nonlinear　relationships.　It　is　found　that　the　deviations　of　both　the　training　and　the　predicting　data　obtained　by　SVM　model　are　much　smaller　than　those　by　ANN　models.　Hence,　SVM　model　can　be　used　as　an　efficient　approach　to　optimize　fabrication　conditions　of　homemade　VC-co-VAc-OH　microfiltration　membrane.