Institution profile

Federal University of Technology - Paraná

Academic institutionsouthamerica · br
Official website
Research library34linked papers
Opportunities0open roles
Selected work

Representative Papers

Spectroscopy Analysis with Machine Learning Regression for the Quantification of Carbon and Nitrogen Contents in Inceptisol and Oxisol Soil Types: Comparing Different Preprocessing and Validation methods as well as Feature Importance

Jul 01, 2026

Traditional methods for soil carbon and nitrogen analysis are time-consuming, costly, and destructive, making them unsuitable for the rapid, non-destructive requirements of modern agriculture. This study addresses this limitation by integrating near-infrared spectroscopy with machine learning to develop an innovative stacking ensemble model tailored for Inceptisols and Oxisols. The proposed approach combines Savitzky–Golay filtering, NIPALS-Huber robust outlier removal, Kennard–Stone sample partitioning, and base learners including PLS, SVR, and Ridge regression, fused via a linear meta-learner. It achieves stable predictive performance with RPD > 2.0 and minimal overfitting across both soil types, significantly outperforming conventional techniques. Furthermore, the work elucidates how soil type influences model generalizability, offering a reliable foundation for in-field, rapid decision-making in precision agriculture.

0 citationsRead paper
Recent publications

Latest Papers

Spectroscopy Analysis with Machine Learning Regression for the Quantification of Carbon and Nitrogen Contents in Inceptisol and Oxisol Soil Types: Comparing Different Preprocessing and Validation methods as well as Feature Importance

Jul 01, 2026

Traditional methods for soil carbon and nitrogen analysis are time-consuming, costly, and destructive, making them unsuitable for the rapid, non-destructive requirements of modern agriculture. This study addresses this limitation by integrating near-infrared spectroscopy with machine learning to develop an innovative stacking ensemble model tailored for Inceptisols and Oxisols. The proposed approach combines Savitzky–Golay filtering, NIPALS-Huber robust outlier removal, Kennard–Stone sample partitioning, and base learners including PLS, SVR, and Ridge regression, fused via a linear meta-learner. It achieves stable predictive performance with RPD > 2.0 and minimal overfitting across both soil types, significantly outperforming conventional techniques. Furthermore, the work elucidates how soil type influences model generalizability, offering a reliable foundation for in-field, rapid decision-making in precision agriculture.

0 citationsRead paper