In the last few years, many efforts have been devoted to the factors controlling the distribution of CMs in karst bauxites, residual deposits hosted in carbonate rocks. Most of these efforts regard Mediterranean-type karst bauxite deposits of Cretaceous age occurring in southern Italy. Further, there is an increasing interest in assessing the usefulness of machine learning applications devoted to geochemically based datasets. With this in mind, we explored a data-centric machine learning arrangement aiming to find the proper input, limited to Al2O3, Fe2O3, TiO2, and SiO2, the most abundant major oxides occurring in these ores, for predicting the HREE distribution in southern Italy karst bauxite deposits. Among the machine learning techniques used, Artificial Neural Network (ANN), Support Vector Machine (SVR), Random Forest (RF) and Extreme Gradient Boosting (XGBoost) are those that effectively predict HREE concentrations. A predictive model based on just Al2O3, Fe2O3, and SiO2, is one conducing at the worst performance impact suggesting that TiO2 is a relevant input variable in order to predict HREE concentrations in considered karst bauxite deposits. The XGBoost model was found to deliver the highest accuracy in predicting HREE for the validation data records (R2 ~ 0.830, RMSE~7.299, MAE ~ 5.091). Moreover, Fe2O3 is the highest correlated input variable with the output variable and is a significant predictor in our model suggesting iron oxyhydroxides play a relevant role in distributing HREE, likely through a scavenging mechanism at the expense of soil solutions. A further step of our research will involve comprehensive cross-validation studies across multiple areas where Mediterranean-type karst bauxite deposits occur, thus providing a thorough assessment of the model's performance. By addressing these tasks and exploring avenues for improvement, the data-centric approach can advance its potential as a cheap and fast technique to perform a preliminary economic evaluation of potentially HREE abundance, as well as other CMs, in karst bauxite ores benefiting applications reliant on these critical resources.

Data-centric approach for predicting critical metals distribution: Heavy rareearth elements in cretaceous Mediterranean-type karst bauxite deposits,southern Italy

Roberto Buccione
Investigation
;
Giovanni Mongelli
Supervision
2023-01-01

Abstract

In the last few years, many efforts have been devoted to the factors controlling the distribution of CMs in karst bauxites, residual deposits hosted in carbonate rocks. Most of these efforts regard Mediterranean-type karst bauxite deposits of Cretaceous age occurring in southern Italy. Further, there is an increasing interest in assessing the usefulness of machine learning applications devoted to geochemically based datasets. With this in mind, we explored a data-centric machine learning arrangement aiming to find the proper input, limited to Al2O3, Fe2O3, TiO2, and SiO2, the most abundant major oxides occurring in these ores, for predicting the HREE distribution in southern Italy karst bauxite deposits. Among the machine learning techniques used, Artificial Neural Network (ANN), Support Vector Machine (SVR), Random Forest (RF) and Extreme Gradient Boosting (XGBoost) are those that effectively predict HREE concentrations. A predictive model based on just Al2O3, Fe2O3, and SiO2, is one conducing at the worst performance impact suggesting that TiO2 is a relevant input variable in order to predict HREE concentrations in considered karst bauxite deposits. The XGBoost model was found to deliver the highest accuracy in predicting HREE for the validation data records (R2 ~ 0.830, RMSE~7.299, MAE ~ 5.091). Moreover, Fe2O3 is the highest correlated input variable with the output variable and is a significant predictor in our model suggesting iron oxyhydroxides play a relevant role in distributing HREE, likely through a scavenging mechanism at the expense of soil solutions. A further step of our research will involve comprehensive cross-validation studies across multiple areas where Mediterranean-type karst bauxite deposits occur, thus providing a thorough assessment of the model's performance. By addressing these tasks and exploring avenues for improvement, the data-centric approach can advance its potential as a cheap and fast technique to perform a preliminary economic evaluation of potentially HREE abundance, as well as other CMs, in karst bauxite ores benefiting applications reliant on these critical resources.
2023
File in questo prodotto:
File Dimensione Formato  
pagination_CHEMER_126026.pdf

solo utenti autorizzati

Licenza: Versione editoriale
Dimensione 1.18 MB
Formato Adobe PDF
1.18 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11563/171455
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? ND
social impact