Improved prediction of soil properties with Multi-target Stacked Generalisation on EDXRF spectra

摘要: Machine Learning (ML) algorithms have been used for assessing soil quality parameters along with non-destructive methodologies. Among spectroscopic analytical methodologies, energy dispersive X-ray fluorescence (EDXRF) is one of the more quick, environmentally friendly and less expensive when compared to conventional methods. However, some challenges in EDXRF spectral data analysis still demand efficient methods capable providing accurate outcomes. Using Multi-target Regression (MTR) methods, multiple can be predicted, also taking advantage inter-correlated overall predictive performance improved. In this study, we proposed Stacked Generalisation (MTSG), a novel MTR method relying on learning from different regressors arranged stacking structure boosted outcome. We MTSG 5 predicting 10 fertility. Random Forest Support Vector (with linear radial kernels) were as embedded into each method. Results showed superiority over Single-target (the traditional ML method), reducing error parameters. Particularly, obtained lowest phosphorus, total organic carbon cation exchange capacity. When observing relative kernel, prediction base saturation percentage was improved 19%. Finally, able reduce average 0.67 (single-target) 0.64 analysing all targets, representing global improvement 4.48%.