Multivariate Time Series Forecasting with Transfer Entropy Graph

Abstract

Multivariate Time Series (MTS) forecasting is an essential problem in many fields. Accurate forecasting results can effectively help in making decisions. To date, many MTS forecasting methods have been proposed and widely applied. However, these methods assume that the predicted value of a single variable is affected by all other variables, ignoring the causal relationship among variables. To address the above issue, we propose a novel end-to-end deep learning model, termed graph neural network with neural Granger causality, namely CauGNN, in this paper. To characterize the causal information among variables, we introduce the neural Granger causality graph in our model. Each variable is regarded as a graph node, and each edge represents the casual relationship between variables. In addition, convolutional neural network filters with different perception scales are used for time series feature extraction, to generate the feature of each node. Finally, the graph neural network is adopted to tackle the forecasting problem of the graph structure generated by the MTS. Three benchmark datasets from the real world are used to evaluate the proposed CauGNN, and comprehensive experiments show that the proposed method achieves state-of-the-art results in the MTS forecasting task.

Anomaly Detection of Wind Turbine Driveline Based on Sequence Decomposition Interactive Network

“…Each variable is regarded as a graph node and each edge means the casual relationship. It is confirmed for multivariate time series forecasting in the dataset of appliances energy consumption [20].…”

Algal Bloom Ties: Spreading Network Inference and Extreme Eco-Environmental Feedback

“…The unbounded causality matrix, or more precisely the predictive causality matrix TE unconstrained to any prediction of biodiversity patterns as in [ 42 ], based on calculated TEs without the optimization of and predictive environmental factors of ecological patterns in an optimal information flow perspective, can be constructed as follows: where is indeed a difference of transfer entropies as in the transfer entropy graph neural network model (TEGNN) (originally developed by Duan et al [ 52 ] and applied to algal blooms by [ 51 ]) in contrast to the optimal information flow model (OIF) originally developed by Li and Convertino [ 20 ]. For each year, two networks were constructed with each defined by an underlying matrix of transfer entropy differences .…”

A recommender system-using novel deep network collaborative filtering

“…Removed nonlinear activations and feature transformations from the network architecture to address graph convolutional networks [15]. They uses GNN to handle a variety of characteristics and created a component to look at the relationship between potential neighbor nodes [16]. Object interactions were captured in using a self-attentive graph neural network and a soft attention method [17].…”

“…Here 𝜏 and 𝜗 are considered as the hyper-parameter so that the denominator is not 0. Based on the values of𝑝𝑟𝑒𝑑(𝑆 𝑓 |𝑆 𝑡,𝑗 ) shown by (16) Based on the results considered 𝑝𝑟𝑒𝑑(𝑡) and 𝑝𝑟𝑒𝑑(𝑡) the value of 𝑝𝑟𝑒𝑑(𝑆 𝑓 |𝑆 𝑡,𝑗 ) is re-evaluated by (15). The evaluation is carried out above until the 𝑝𝑟𝑒𝑑(𝑆 𝑓 |𝑆 𝑡,𝑗 ) that reaches the community to achieve the large value for 𝑝𝑟𝑒𝑑(𝑆 𝑓 |𝑆 𝑡,𝑗 ) as the final community that links the 𝑆 𝑡,𝑗 .…”