Abstract:Spatial prediction models hold significant application value in fields such as environmental science, economic development, and geological exploration. With advancements in deep learning technology, graph neural networks (GNNs) offer a powerful and scalable solution for spatial data modeling. In these models, each spatial location is represented as a vertex, with the explanatory variables at each point converted into vertex features, and the vertex label corresponding to the target value at that point. GNNs utilize this representation to predict vertex labels, thereby performing spatial prediction. However, the residuals of GNN predictions are found to still exhibit spatial autocorrelation, indicating that traditional GNNs only achieve suboptimal results in terms of capturing complex spatial relationships. In this paper, we propose a two-stage spatial prediction method called Location Embedded Graph Neural Networks-Residual Neural Processes (LEGNN-RNP) to address this challenge. In the first stage, we employ LEGNNs, a new framework that integrates location features into GNNs through an attention mechanism, enhancing the learning of complex spatial models by considering both spatial context and attribute features. In the second stage, we model the residuals from LEGNN predictions to further extract spatial patterns from the data. Specifically, we introduce RNP, a neural process-based approach to model the Gaussian Process (GP) distribution of residuals. The distribution parameters (i.e., mean and covariance) are parameterized by a neural network, where the mean estimates the residual and the variance quantifies the prediction uncertainty. Experiments on four datasets demonstrate that our proposed two-stage method achieves state-of-the-art results by effectively extracting spatial relationships and significantly improving spatial prediction accuracy.

Generalized spatial linear regression model for spatial prediction

Local Linear M‐estimation in Non‐parametric Spatial Regression

A Generalized Spatial Autoregressive Neural Network Method for Three-Dimensional Spatial Interpolation

Geographically weighted regression with the integration of machine learning for spatial prediction

A Spatiotemporal Regression-Kriging Model for Space-Time Interpolation: a Case Study of Chlorophyll-a Prediction in the Coastal Areas of Zhejiang, China.

Variable Selection for Spatial Nonparametric Regression

Combining Regression Kriging With Machine Learning Mapping for Spatial Variable Estimation

Mixed Geographically Weighted Regression Model-Based Small Area Estimation

The Geostatistical Framework for Spatial Prediction

Variable Selection of Spatial Logistic Autoregressive Model with Linear Constraints

Spatial Autologistic Model with Generalized Dependent Parameter

Indicator and Multivariate Geostatistics for Spatial Prediction

Variable Selection for Spatial Logistic Autoregressive Models

Local Walsh-average Regression for Spatial Autoregression Single Index Varying Coefficient Models

Nonparametric Spatial Regression with Spatial Autoregressive Error Structure

A Two-Stage Spatial Prediction Modeling Approach Based on Graph Neural Networks and Neural Processes

Uniform expression of variogram nested model and parameter estimation in spatial prediction of soil properties

Generalized Spatial Regression with Differential Regularization

Spatial homogeneity pursuit of regression coefficients for large datasets

Nonparametric spatial autoregressive model using deep neural networks

Geospatial Least Squares Support Vector Regression Fused with Spatial Weight Matrix