Abstract:Time series forecasting and spatiotemporal kriging are the two most important tasks in spatiotemporal data analysis. Recent research on graph neural networks has made substantial progress in time series forecasting, while little attention has been paid to the kriging problem -- recovering signals for unsampled locations/sensors. Most existing scalable kriging methods (e.g., matrix/tensor completion) are transductive, and thus full retraining is required when we have a new sensor to interpolate. In this paper, we develop an Inductive Graph Neural Network Kriging (IGNNK) model to recover data for unsampled sensors on a network/graph structure. To generalize the effect of distance and reachability, we generate random subgraphs as samples and reconstruct the corresponding adjacency matrix for each sample. By reconstructing all signals on each sample subgraph, IGNNK can effectively learn the spatial message passing mechanism. Empirical results on several real-world spatiotemporal datasets demonstrate the effectiveness of our model. In addition, we also find that the learned model can be successfully transferred to the same type of kriging tasks on an unseen dataset. Our results show that: 1) GNN is an efficient and effective tool for spatial kriging; 2) inductive GNNs can be trained using dynamic adjacency matrices; 3) a trained model can be transferred to new graph structures and 4) IGNNK can be used to generate virtual sensors.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the application of real - time spatial interpolation (i.e., kriging) on dynamic network structures. Specifically, the authors developed an inductive model based on graph neural networks (GNN) - Inductive Graph Neural Network Kriging (IGNNK), which is used to infer signals at unsampled locations/sensors from the data of known sensors. ### Problem Background 1. **Spatio - temporal Kriging Problem**: The goal of spatio - temporal Kriging is to interpolate signals at unsampled locations from the position signals of known sensors. This task is crucial for improving the resolution and accuracy of spatio - temporal data and can be applied in fields such as travel time estimation and disaster assessment. 2. **Limitations of Existing Methods**: - **Gaussian Process Regression (GP)**: Although it is suitable for spatio - temporal Kriging problems in the Euclidean domain, it has high computational complexity, is difficult to handle large - scale data sets, and is difficult to model network systems. - **Matrix/Tensor Completion**: These methods are essentially transductive. When new sensors are introduced, the entire model needs to be retrained and cannot be directly applied to new graph structures, and low - rank schemes are difficult to adapt to dynamically changing graph structures. ### Solutions Proposed in the Paper To solve the above problems, the paper proposes the Inductive Graph Neural Network Kriging (IGNNK) model. The main contributions of IGNNK include: 1. **Inductive Ability**: Different from existing transductive methods, IGNNK has inductive ability, can learn the message - passing mechanism, and generalize it to unseen nodes or graph structures. 2. **Random Sub - graph Generation**: Generate sample sub - graphs by randomly selecting subsets of nodes and reconstruct the adjacency matrix of each sub - graph. This method enables the model to better learn the spatial message - passing mechanism and improve the generalization ability for unseen nodes. 3. **Dynamic Adjacency Matrix**: IGNNK can use a dynamic adjacency matrix during the training process, thereby adapting to dynamically changing graph structures. 4. **Virtual Sensor Generation**: IGNNK can be used to generate virtual sensors, thereby reducing the number of actual sensors and reducing operation and maintenance costs. ### Experimental Results The paper conducted experiments on five real - world spatio - temporal data sets, including traffic speed data sets (METR - LA, SeData, PeMS - Bay), a solar energy output data set (NREL), and a monthly precipitation data set (USHCN). The experimental results show that IGNNK achieved the best performance on almost all data sets, especially on traffic speed data sets. In addition, IGNNK also demonstrated good transfer learning ability and was able to achieve good interpolation effects on unseen data sets. ### Summary The paper solves the key challenges in spatio - temporal Kriging problems through the IGNNK model, especially how to perform effective signal interpolation on dynamic network structures. IGNNK not only improves the accuracy and efficiency of interpolation but also demonstrates strong generalization and transfer learning abilities.

Inductive Graph Neural Networks for Spatiotemporal Kriging

INCREASE: Inductive Graph Representation Learning for Spatio-Temporal Kriging

Spatial Aggregation and Temporal Convolution Networks for Real-time Kriging

STONet: A Neural-Operator-Driven Spatio-temporal Network

KITS: Inductive Spatio-Temporal Kriging with Increment Training Strategy

Inductive Spatial Temporal Prediction Under Data Drift with Informative Graph Neural Network

Scalable Spatiotemporal Graph Neural Networks

Spatial-Temporal Aware Inductive Graph Neural Network for C-ITS Data Recovery

It is all Connected: A New Graph Formulation for Spatio-Temporal Forecasting

Spatial-Temporal Interactive Graph Neural Network for Traffic Forecasting.

Inference of Spatio-Temporal Functions over Graphs via Multi-Kernel Kriged Kalman Filtering

Isometric Graph Neural Networks

Spatial-Temporal Knowledge Graph Network for Event Prediction

Unboxing the graph: Neural Relational Inference for Mobility Prediction

Graph Neural Network for spatiotemporal data: methods and applications

Graph convolutional networks for spatial interpolation of correlated data

Dynamic Causal Explanation Based Diffusion-Variational Graph Neural Network for Spatiotemporal Forecasting

DKNN: deep kriging neural network for interpretable geospatial interpolation

STGIN: A Spatial Temporal Graph-Informer Network for Long Sequence Traffic Speed Forecasting

Graph construction on complex spatiotemporal data for enhancing graph neural network-based approaches

Spatial-Temporal Graph Deviation Neural Networks for Anomaly Detection in Time Series