基于多证据融合模型的洪水预测研究  

Study of flood prediction based on multi-evidential fusion model

作　　者：吴蔚[1] 张海波[2] 王道席[3] 

机构地区：[1]浙江大学,浙江杭州310027 [2]浙江水利水电专科学校,浙江杭州310018 [3]水利部黄河水利委员会,河南郑州450003

出　　处：《水力发电》2005年第12期22-24,共3页Water Power

基　　金：国家重点基础研究发展规划(973)资助项目(G19990436);浙江省水利厅科技项目(RC0509)

摘　　要：针对单一模型解决洪水预测问题存在的算法复杂度高、分类准确率低等问题,提出了BP神经网络联合与DS证据推理相融合的模型,不仅实现了多个领域不同层次的全部主?客观证据的特征级融合,还实现了多个模型的优势互补。通过实验对该方法和传统的单一神经网络方法比较得出,主?客观证据融合方法不仅提高了4.4%的分类精度,还降低了算法的时间和复杂度。There are problems such as high complexity of algorithms and low accuracy rate of classifications lie in the flood prediction using single models. To solve these problems, the paper presents a modal by combining BP neural network and DS evidential reasoning, which not only achieves the feature level fusion of all subjective and objective evidences in various domains and layers, but also makes distinct models complement each other. Comparing with traditional single neu,＇al network by the experiment, this method improves classification precision by 4.4 percent and reduces the time complexity of algorithm.

关 键 词：信息融合 洪水预测 神经网络 DS证据理论 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]