GPS elevation time series prediction model based on wavelet denoising technique and neural network
-
摘要: 全球定位系统(GPS)时序数据预测的工作中发现,通常时序数据中含有的噪声会干扰数据预测的结果.为了降低时序数据中噪声对预测结果的负面影响,将提升小波阈值降噪技术和长短期记忆(LSTM)神经网络相结合,实现一种GPS时序数据降噪预测模型.该模型在预测之前首先利用提升小波与平滑阈值函数对GPS时序数据中的噪声进行剥离,然后构建多层LSTM神经网络对时序数据进行单步预测.通过实验与多种时间序列预测模型进行对比,结果表明所提出的LSTM预测模型对GPS时间序列的预测具有较强的适用性和较高的准确性.Abstract: In the work of GPS time series data prediction, it is found that the noise contained in the time series data usually interferes with the results of data prediction. In order to minimize the negative impact of noise on prediction results in time series data as much as possible, a GPS time series data denoising prediction model is realized by combining lifting wavelet threshold denoising technology with long Short-Term Memory (LSTM). The model first uses lifting wavelet and smoothing threshold function to remove noise from GPS time series data before prediction, and then constructs a multi-layer LSTM neural network to predict time series data in one step. Compared with many time series prediction models, the results show that the proposed LSTM prediction model has strong applicability and higher accuracy for GPS time series prediction.
-
[1] 张晗, 王霞. 基于小波分解的网络流量时间序列建模与预测[J]. 计算机应用研究, 2012, 29(8):3134-3136. [2] 王书芹, 华钢, 郝国生,等. 基于灰狼优化算法的长短期记忆网络在时间序列预测中的应用[J]. 中国科技论文, 2017,12(20):2309-2314. [3] 吴志华, 丁杨斌, 申功勋. 非平稳时序分析在GPS伪距观测值建模中的应用[J]. 系统仿真学报, 2008, 20(16):4252-4254,4260. [4] 魏鸿浩, 贾云峰. 一种非平稳非线性频谱占用度时间序列分析方法[J]. 电子学报, 2017, 45(8):2026-2030. [5] 龙勇, 苏振宇, 汪於. 基于季节调整和BP神经网络的月度负荷预测[J]. 系统工程理论与实践, 2018, 38(4):1052-1060. [6] COULIBALY P, BALDWIN C K. Nonstationary hydrological time series forecasting using nonlinear dynamic methods[J]. Journal of Hydrology, 2005, 307(1-4):164-174. DOI: 10.1016/j.jhydrol.2004.10.008. [7] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8):1735-1780.DOI:10.1162/neco.1997.9.8.1735·Source: PubMed. [8] GRAVES A. Supervised sequence labelling with recurrent neural networks[M]. Springer Berlin Heidelberg, 2012. DOI: 10.1007/978-3-642-24797-2. [9] LOPEZ E, VALLE C, ALLENDE H, et al. Wind power forecasting based on echo state networks and long short-term memory[J]. Energies, 2018, 11(3):526. DOI: 10.3390/en11030526. [10] 李基武, 符强, 孙希延,等. 一种改进扩展卡尔曼滤波算法的GPS多径抑制应用[J]. 测绘科学, 2018(6):167-172. [11] 江虹, 苏阳. 一种改进的小波阈值函数去噪方法[J]. 激光与红外, 2016, 46(1):119-122. [12] MALLAT S. A theory for multiresolution vision model applied to astronomical images[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 1989(40):495-520. [13] CAO X, ZHANG Z. Method of radar signal de-noising based on lifting wavelet improved threshold[J]. Computer Engineering and Applications, 2012, 48(14):143-147. [14] 周风波, 李长庚, 朱红求. 基于提升小波变换的阈值改进去噪算法在紫外可见光谱中的研究[J]. 光谱学与光谱分析, 2018,38(2):506-510. [15] 李明迅, 孟相如, 袁荣坤,等. 融合提升小波降噪和LSSVM的网络流量在线预测[J]. 计算机应用, 2012, 32(2):340-346,346. [16] 方志军, 石恒麟, 毛微微. 基于小波降噪的RLS算法在股票预测中的应用[C]//International Conference on Management Science and Engineering. 2010. [17] DONOHO D L, JOHNSTONE I M. Ideal spatial adaptation by wavelet shrinkage[J]. Biometrika, 1994, 81(3):425-455. DOI: 10.1093/biomet/81.3.425. [18] DONOOHO D L. De-noising by soft-thresholding[J]. IEEE Transactions on Information Theory,1995, 41(3):613-627. DOI: 10.1109/18.382009. [19] 符养. 中国大陆现今地壳形变与GPS坐标时间序列分析[D].上海:中国科学院研究生院(上海天文台), 2002. [20] 张鹏, 蒋志浩, 秘金钟,等. 我国GPS跟踪站数据处理与时间序列特征分析[J]. 武汉大学学报(信息科学版), 2007, 32(3):251-254. [21] 张剑, 屈丹, 李真. 基于词向量特征的循环神经网络语言模型[J]. 模式识别与人工智能, 2015, 28(4): 299-305.
点击查看大图
计量
- 文章访问数: 478
- HTML全文浏览量: 89
- PDF下载量: 105
- 被引次数: 0