The　applicability　of　model　predictive　accuracy　and　the　selection　of　suitable　model　parameters　for　the　membrane　flux　prediction　and　prediction　of　the　filtration　characteristics　of　dead-end　filtration　systems　were　assessed.　In　determining　the　model　predictive　accuracy　using　model　parameters,　the　selection　of　suitable　model　parameters　is　of　great　significance.　A　series　of　experiments　were　conducted.　The　relative　deviations,　average　deviations,　and　root　mean　square　between　the　experimental　and　predicted　values　were　calculated　to　evaluate　predictive　accuracy.　The　results　showed　that　the　model　parameter　has　changed　nonlinearly　with　operating　conditions.　Particularly,　in　the　membrane　flux　prediction　model,　the　time-dependent　character　of　the　model　parameter　can　result　in　a　considerable　error.　However,　model　prediction　accuracy　can　be　improved　indisputably　and　the　model　will　get　higher　predictive　accuracy　than　that　of　other　predictive　models　if　the　constant　membrane　fouling　index　in　the　flux　prediction　model　is　either　replaced　by　the　average　membrane　fouling　index　or　it　takes　a　composite　parameter　with　free　combination　of　the　initial　membrane　fouling　index,　meantime,　minimum　and　the　maximum　of　the　model　parameter.　These　results　give　us　a　possibility　to　see　the　panorama　of　the　variation　of　the　model　parameters　in　different　situations　and　an　idea　to　establish　the　foundation　for　building　mathematical　models.