GTCNNWR: Geographically and Temporally Convolutional Neural Network Weighted Regression for Accurate Spatiotemporal Modelling on Uneven Data

This repository provides the official implementation of the Geographically and Temporally Convolutional Neural Network Weighted Regression (GTCNNWR) model, which addresses spatiotemporal non-stationarity through a novel architecture combining global spatiotemporal division (GSPG/GTPA) with specialized convolutional neural networks and weighted neural networks. The included code features model definitions, training pipelines, and evaluation scripts that demonstrate the model's enhanced robustness for unevenly distributed spatiotemporal data, as validated through both simulated and environmental datasets.