Python scripts and Dataset for a PU-learning framework for predicting agriculture terrace suitability: an application for the Troodos Mountains, Cyprus

This repository contains the rasterized dataset and Python scripts used to develop a machine learning framework for predicting agricultural terrace suitability in the Troodos Mountains, Cyprus. The dataset includes environmental, regulatory, and socio-economic variables, while the scripts implement Positive and Unlabeled (PU) learning using XGBoost models. All datasets can also be visualized via the Google Earth Engine at the following link: https://ee-ameena.projects.earthengine.app/view/land-suitability-analysis Note that original crop plot data, which have data-sharing restrictions, are not shared. The rasterized dataset only contains binary data on presence (1) or absence of a crop plot (0).