| 基于NARX神经网络的港口集装箱吞吐量预测 |
| |
| 引用本文： | 范莹莹,余思勤.基于NARX神经网络的港口集装箱吞吐量预测[J].上海海事大学学报,2015,36(4):1-5. |
| |
| 作者姓名： | 范莹莹 余思勤 |
| |
| 作者单位： | 上海海事大学 经济管理学院,上海海事大学 经济管理学院 |
| |
| 基金项目： | 国家自然科学基金（61273042） |
| |
| 摘 要： | 为对港口集装箱吞吐量进行科学预测,采用带外生变量的非线性自回归(NARX)模型对上海港的集装箱吞吐量进行预测.通过主成分分析法对港口吞吐量影响因子进行相关性分析,将筛选出的GDP作为外部输入因子引入NARX模型.实证分析发现,引入GDP的NARX神经网络模型对具有非线性特征的集装箱吞吐量数据有良好的映射逼近性.训练后的网络误差小且拟合度高,具有良好的泛化能力,预测性能较好.
|
| 关 键 词： | NARX神经网络 集装箱吞吐量 主成分分析 动态预测 |
| 收稿时间： | 2015/5/14 0:00:00 |
| 修稿时间： | 2015/6/11 0:00:00 |
Port container throughput forecast based on NARX neural network |
| |
| Institution: | School of Economics and Management,Shanghai Maritime University |
| |
| Abstract: | In order to predict port container throughput scientifically, a nonlinear auto-regressive model with exogenous inputs (NARX) model is used to forecast the container throughput of Shanghai Port. After analyzing the correlation of port throughput impact factors by the principal component analysis, GDP is selected as an exogenous input of the NARX model. Through the empirical analysis, it is found that the NARX neural network model with GDP is of good approximation of mapping to the port container throughput data with nonlinear characteristics. The trained network is of small error, high fitting degree, good generalization ability and good prediction performance. |
| |
| Keywords: | NARX neural network container throughput principal component analysis dynamic forecast |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《上海海事大学学报》浏览原始摘要信息 |
| 点击此处可从《上海海事大学学报》下载免费的PDF全文 |
