Machine Learning Predictive Models for Phosphate Exploration

Abstract

Machine learning methods particularly, partial least squares regressions (PLSR) and Support Vector Machine (SVM), were conducted in this study for the abundance prediction of phosphate minerals. Abundances of fluorapatite and dolomite minerals were predicted using VNIR-SWIR hyperspectral reflectance and abundance results issued from X-Ray Diffraction analysis. The performance statistics of the generated PLSR models, for both fluorapatite and dolomite minerals, revealed by the X-Ray Diffraction analysis, were calculated. According to the results, the short-wave infrared Short Wave InfraRed SWIR region has been shown to be the most important for the prediction of dolomite and fluorapatite contents. The PLSR model, developed for the fluorapatite content prediction, has shown an interesting performance compared to the once based on the SVM method. For the dolomite, the SVM model shoed the better results.