基于图卷机网络的多元时间序列预测模型及方法
By using a multivariate time series forecasting model based on a graph roll machine network, the problem of complex dependencies in multivariate time series forecasting is solved, achieving efficient capture of temporal and spatial dependencies and improving forecast accuracy and speed.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NORTHWEST UNIV
- Filing Date
- 2023-03-14
- Publication Date
- 2026-07-17
AI Technical Summary
In multivariate time series forecasting, existing technologies struggle to effectively capture the dependencies between time and spatial scales, resulting in insufficient forecast accuracy. In particular, recurrent neural networks (RNNs) suffer from the vanishing gradient problem and do not support parallel computing, which affects both forecast speed and accuracy.
A multivariate time series prediction model based on graph convolutional neural network is adopted, including a correlation layer, a graph filtering layer, a time series processing layer, and a prediction layer. The graph filter is constructed by GRU module, Attention module, Laplace module, and Cheb module. Combined with dual time series processing layer and fully connected neural network, parallel processing and efficient prediction of multivariate time series are achieved.
It improves the accuracy and speed of multivariate time series forecasting, effectively captures temporal and spatial dependencies, and enhances forecast precision and efficiency.
Smart Images
Figure CN116485005B_ABST