摘要
针对赤潮爆发受多因素影响且影响因素间存在相关性的特性,提出了一种基于主成分分析改进的PSO-BP神经网络预测模型(PCA-PSO-BP)。在充分利用原始观测信息的前提下,采用主成分分析消除变量间的相关性,从而减少网络输入节点数,简化网络模型,同时结合粒子群优化算法初始化网络初始权值和阈值,建立高精度PCA-PSO-BP神经网络预测模型。结合赤潮监测实例数据,分别与传统BP神经网络、PCA-BP神经网络、PSO-BP神经网络预测模型进行对比,结果表明采用PCA-PSO-BP神经网络预测赤潮具有一定的可行性,可以提高预测模型精度。
Aiming at the characteristics that red tide outbreak was affected by many factors and there was correlation among them,PSO-BP prediction model based on principal component analysis(PCA-PSO-BP)is proposed in this paper.Under the premise of making full use of the original information,the principal component analysis is used to eliminate the correlation between variables.Then the network model is simplified by reducing the number of input nodes.Finally,the initial weights and thresholds of the network are initialized by particle swarm optimization algorithm to form a high-precision PCA-PSO-BP neural network prediction model.Combining with the actual data of red tide monitoring,the prediction models of BP neural network,PCA-BP neural network and PSO-BP neural network are compared.The results showed that the prediction model of red tide using PCA-PSO-BP neural network is feasible,which improves the accuracy of prediction model.
作者
徐定建
XU Dingjian(Chongqing Survey Institute,Chongqing 401121,China)
出处
《测绘通报》
CSCD
北大核心
2021年第S02期234-240,共7页
Bulletin of Surveying and Mapping
关键词
赤潮爆发
主成分分析
粒子群优化算法
神经网络
预测
red tide outbreak
principal component analysis
particle swarm optimization algorithm
neural network
prediction
