@article{article_1702675, title={Hight-Precision 3D Coordinate Transformation Using XGBoost Regression: A Machine Learning Approach for Geospatial Data}, journal={Turkish Journal of Engineering}, volume={10}, pages={91–102}, year={2025}, DOI={10.31127/tuje.1702675}, author={Mohamed, Hossamaldeen and Hassan, Abubakr and Kamaleldeen, Maab}, keywords={XGBoost, 3D Coordinate Prediction, Geospatial Transformation, Regression Modeling, Machine Learning, RMSE, R² Score}, abstract={Traditional methods for 3D coordinate transformation often struggle with complex mathematical computations. This study presents a machine learning approach using Extreme Gradient Boosting (XGBoost) to achieve high-precision coordinate transformations between measurement systems. We developed three specialized XGBoost models (X, Y, Z axes) that learn the transformation rules directly from data, eliminating the need for predefined mathematical models. The framework processed raw coordinate measurements through careful data cleaning and splitting (80% training, 20% testing), intentionally avoiding normalization to preserve transformation relationships. Results demonstrated exceptional transformation accuracy, with R² scores of 0.9999 (X), 0.9996 (Y), and 0.9975 (Z), and RMSE values as low as 0.185 units. Error analysis showed maximum deviations under 1.5 units across all axes, while 3D visualization confirmed the model’s ability to maintain geometric relationships during transformation. The independent axis modeling approach proved particularly effective for coordinate system conversions, capturing axis-specific transformation characteristics without cross-contamination. This work establishes XGBoost as a powerful alternative to conventional transformation methods, offering superior accuracy for applications in geodesy, photogrammetry, and CAD systems. Future enhancements could incorporate hybrid models that combine the strengths of parametric transformations with machine learning refinements.}, number={1}, publisher={Murat YAKAR}