| 两种样本输入方式下基于GRNN的日长变化预报结果的比较 |
| |
| 引用本文： | 张晓红,王琪洁,朱建军,张昊.两种样本输入方式下基于GRNN的日长变化预报结果的比较[J].中国科学院上海天文台年刊,2011(0). |
| |
| 作者姓名： | 张晓红 王琪洁 朱建军 张昊 |
| |
| 作者单位： | 中南大学地球科学与信息物理学院; |
| |
| 基金项目： | 国家自然科学基金委员会与中国科学院天文联合基金(No.10878026) |
| |
| 摘 要： | 针对广义回归神经网络用于日长变化预报过程中,样本的输入方式对预报结果的影响进行了研究。采用2种输入方式:即样本按不同跨度输入以及按连续输入,对日长变化进行预报。最终证明不同的样本输入方式对日长变化预报精度的影响较大,样本按跨度输入在超短期预报中预报精度较高,样本采用连续输入的方式在短期和中期预报中预报精度较高。
|
| 关 键 词： | 广义回归神经网络 日长变化预报 输入方式 相关性 |
Comparison of Results of Forecasting LOD Based on GRNN with two Input Architectures |
| |
| ZHANG Xiao-hong,WANG Qi-jie,ZHU Jian-jun,ZHANG Hao.Comparison of Results of Forecasting LOD Based on GRNN with two Input Architectures[J].Annals of Shanghai Observatory Academia Sinica,2011(0). |
| |
| Authors: | ZHANG Xiao-hong WANG Qi-jie ZHU Jian-jun ZHANG Hao |
| |
| Institution: | ZHANG Xiao-hong,WANG Qi-jie,ZHU Jian-jun,ZHANG Hao (School of Geosciences and Info-Physics,Central South University,Changsha 410083) |
| |
| Abstract: | A comparison of the result of two input architectures,sample in different lead time and sample in sequence,in forecasting LOD by General Regression Neural Network (GRNN) model is presented.Results show that two different input architectures have obvious different influence on accuracy of the LOD prediction.The former leads to higher accuracy in ultra short-term prediction,while the latter leads to higher accuracy in short-term and medium-term predictions. |
| |
| Keywords: | General Regression Neural Network(GRNN) prediction of LOD input architecture relativity |
| 本文献已被 CNKI 等数据库收录！ |
