[general]
name=GPBoost Spatial Predictor
qgisMinimumVersion=3.28
qgisMaximumVersion=3.99
description=Spatial prediction with GPBoost, combining tree boosting and Gaussian processes for point observations.
version=1.0.2
author=Jesús Enrique Flores-Riera, Santiago Gallego
email=lab227fca@gmail.com
about=GPBoost Spatial Predictor trains GPBoost models for spatial prediction from point layers with numeric response and covariate fields. It combines tree boosting for nonlinear covariate-response relationships with a Gaussian Process component for spatial residual autocorrelation. External dependency: the Python package gpboost>=1.4.0 must be installed in the same Python environment used by QGIS. The plugin includes an interactive dialog, a QGIS Processing algorithm, model comparison/tuning tools, and English, Spanish, and Portuguese interface labels. Current limitation: when creating prediction rasters from selected covariate fields, covariates are fixed at their median values over the prediction grid; future versions should accept raster covariate layers for fully covariate-varying maps.
homepage=https://github.com/jf-floresriera/GPBoost-Spatial-Predictor#readme
tracker=https://github.com/jf-floresriera/GPBoost-Spatial-Predictor/issues
repository=https://github.com/jf-floresriera/GPBoost-Spatial-Predictor
license=GPL-2.0-or-later
tags=machine learning, spatial prediction, gaussian process, boosting, interpolation, raster
category=Raster
icon=icons/icon.png
experimental=True
deprecated=False
hasProcessingProvider=True
server=False
changelog=1.0.2 - Fixes QVariant/null numeric conversion for QGIS feature attributes and confirms QGIS 3.28-3.99 compatibility only.
    1.0.1 - Publication-preparation release: English default UI, Spanish and Portuguese runtime translations, valid icon path, README, LICENSE, requirements, sample CSV, Processing metadata, safer dependency messaging, and corrected GP covariance handling.
    1.0.0 - Initial GPBoost training, comparison, and Processing prototype.
