摘要
针对港口集装箱吞吐量在时序上的复杂非线性特征,用自相关分析技术分析时间序列的延迟特性,确定神经网络的输入向量维数;用差分法对数据进行预处理,去除缓变部分,并在一定程度上减少随机噪声;引入延时内部反馈,构造基于径向基函数(RBF)网络的动态预测模型,并用大连港集装箱吞吐量实际数据进行了验证.结果表明:该模型拟合精度和预测精度高,具有很好的预测能力和应用价值.
Aiming at the complicated nonlinear characters of the time serial of the port container throughput, the delayed character of time serial is studied by the technique of self-relativity analysis. Based on the analysis result, the input vectors of a neural network are confirmed. With the data preprocessing using differential method to remote slow change and reduce random noise, the RBF dynamic prediction model is constructed based on delay internal feedback. The actual container throughput data of Dalian port prove that the prediction is accurate and precise in fitting and prediction precision.
出处
《大连交通大学学报》
CAS
2008年第4期27-32,共6页
Journal of Dalian Jiaotong University
基金
辽宁省教育厅科学研究计划资助项目(05L034)
关键词
自相关分析
差分法
动态预测
集装箱吞吐量
RBF算法
self-relativity analysis
difference method
dynamic prediction
container throughput
RBF algorithms
作者简介
李季涛（1971-），男，讲师，硕士,E-mail：to_lijitao@hotmail．com．
