Code and data for "Two-stage Variational Mode Decomposition and Support Vector Regression for Streamflow Forecasting"

This data repository contains code and data for the research article “Two-stage Variational Mode Decomposition and Support Vector Regression for Streamflow Forecasting”, which is currently under review for the journal Hydrology and Earth System Sciences (HESS). The underlying data of this study is the monthly runoff data sets (from Jan 1953 to Dec 2018) of Huaxian, Xianyang, and Zhangjiashan stations, Wei River, China, which is organized in "time_series" directory. The unit of measurement is 10^8m^3/s. The fundamental code for decomposing runoff data, deciding input predictors and output target, generating machine learning samples, building Autoregressive moving average (ARIMA), support vector regression (SVR), Backpropagation neural network (BPNN), and Long short-term memory (LSTM) models, and evaluating the model performance is organized in the “tools” directory. The execution code for forecasting different runoff series using different decomposition algorithms (e.g., variational mode decomposition (VMD), ensemble empirical mode decomposition (EEMD), discrete wavelet transform (DWT), Singular spectrum analysis (SSA) or non-decomposition-based (Orig)) are organized in “projects” directory (e.g., “huaxian_vmd/projects/”). To reproduce the results of this paper, follow the instructions given in readme.md. Note that the same results demonstrated in this paper cannot be reproduced but similar results should be reproduced.本数据集仓库包含为研究论文《Two-stage Variational Mode Decomposition and Support Vector Regression for Streamflow Forecasting》（目前正在《水文与地球系统科学》（Hydrology and Earth System Sciences, HESS）期刊审稿中）所编写的代码与配套数据。本研究的基础数据为中国渭河华县、咸阳、张家山水文站1953年1月至2018年12月的月度径流数据集，存储于time_series目录中。其计量单位为10^8立方米每秒。用于径流数据分解、确定输入预测因子与输出目标、生成机器学习样本、构建自回归移动平均（Autoregressive Moving Average, ARIMA）、支持向量回归（Support Vector Regression, SVR）、反向传播神经网络（Backpropagation Neural Network, BPNN）以及长短期记忆网络（Long Short-Term Memory, LSTM）模型，并评估模型性能的基础代码，收纳于tools目录中。使用不同分解算法（例如变分模态分解（Variational Mode Decomposition, VMD）、集合经验模态分解（Ensemble Empirical Mode Decomposition, EEMD）、离散小波变换（Discrete Wavelet Transform, DWT）、奇异谱分析（Singular Spectrum Analysis, SSA）或非分解基准（Orig））开展不同径流序列预报的执行代码，收纳于projects目录中（示例路径：“huaxian_vmd/projects/”）。若需复现本论文的研究结果，请遵循readme.md中给出的操作指引。请注意：无法复现本文中展示的完全一致的结果，但可复现相似结果。