Spatial Regression

Spatial regression aims to model relationships between variables while explicitly accounting for their spatial locations, seeking to understand how these locations influence the relationships and improve prediction accuracy. Current research focuses on developing more efficient and interpretable models, including attention-based architectures and novel algorithms like covariate-distance weighted regression and the modified planar rotator method, which address limitations of traditional geographically weighted regression. These advancements enhance the accuracy and explainability of spatial analyses across diverse fields, from geographic modeling and image analysis to house price prediction and even table structure recognition in document processing. Improved methods are leading to better understanding of spatial heterogeneity and more reliable predictions in various applications.