Application of Machine Learning Algorithms to Classification of Pb–Zn Deposit Types Using LA–ICP–MS Data of Sphalerite
Pb–Zn deposits supply a significant proportion of critical metals, such as In, Ga, Ge, and Co. Due to the growing demand for critical metals, it is urgent to clarify the different types of Pb–Zn deposits to improve exploration. The trace element concentrations of sphalerite can be used to classify t...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-10-01
|
| Series: | Minerals |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2075-163X/12/10/1293 |
| _version_ | 1797471023585558528 |
|---|---|
| author | Guo-Tao Sun Jia-Xi Zhou |
| author_facet | Guo-Tao Sun Jia-Xi Zhou |
| author_sort | Guo-Tao Sun |
| collection | DOAJ |
| description | Pb–Zn deposits supply a significant proportion of critical metals, such as In, Ga, Ge, and Co. Due to the growing demand for critical metals, it is urgent to clarify the different types of Pb–Zn deposits to improve exploration. The trace element concentrations of sphalerite can be used to classify the types of Pb–Zn deposits. However, it is difficult to assess the multivariable system through simple data analysis directly. Here, we collected more than 2200 analyses with 14 elements (Mn, Fe, Co, Ni, Cu, Ga, Ge, Ag, Cd, In, Sn, Sb, Pb, and Bi) from 65 deposits, including 48 analyses from carbonate replacement (CR), 684 analyses from distal magmatic-hydrothermal (DMH), 197 analyses from epithermal, 456 analyses from Mississippi Valley-type (MVT), 199 analyses from sedimentary exhalative (SEDEX), 377 analyses from skarn, and 322 analyses from volcanogenic massive sulfide (VMS) types of Pb–Zn deposits. The critical metals in different types of deposits are summarized. Machine learning algorithms, namely, decision tree (DT), K-nearest neighbors (KNN), naive Bayes (NB), random forest (RF), and support vector machine (SVM), are applied to process and explore the classification. Learning curves show that the DT and RF classifiers are the most suitable for classification. Testing of the DT and RF classifier yielded accuracies of 91.2% and 95.4%, respectively. In the DT classifier, the feature importances of trace elements suggest that Ni (0.22), Mn (0.17), Cd (0.13), Co (0.11), and Fe (0.09) are significant for classification. Furthermore, the visual DT graph shows that the Mn contents of sphalerite allow the division of the seven classes into three groups: (1) depleted in Mn, including MVT and CR types; (2) enriched in Mn, including epithermal, skarn, SEDEX, and VMS deposits; and (3) DMH deposits, which have variable Mn contents. Data mining also reveals that VMS and skarn deposits have distinct Co and Ni contents and that SEDEX and DMH deposits have different Ni and Ge contents. The optimal DT and RF classifiers are deployed at Streamlit cloud workspace. Researchers can select DT or RF classifier and input trace element data of sphalerite to classify the Pb–Zn deposit type. |
| first_indexed | 2024-03-09T19:43:40Z |
| format | Article |
| id | doaj.art-b981f665498b4af49c32e996f72ef82b |
| institution | Directory Open Access Journal |
| issn | 2075-163X |
| language | English |
| last_indexed | 2024-03-09T19:43:40Z |
| publishDate | 2022-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Minerals |
| spelling | doaj.art-b981f665498b4af49c32e996f72ef82b2023-11-24T01:30:13ZengMDPI AGMinerals2075-163X2022-10-011210129310.3390/min12101293Application of Machine Learning Algorithms to Classification of Pb–Zn Deposit Types Using LA–ICP–MS Data of SphaleriteGuo-Tao Sun0Jia-Xi Zhou1State Key Laboratory of Public Big Data, Guizhou University, Guiyang 550025, ChinaSchool of Earth Sciences, Yunnan University, Kunming 650500, ChinaPb–Zn deposits supply a significant proportion of critical metals, such as In, Ga, Ge, and Co. Due to the growing demand for critical metals, it is urgent to clarify the different types of Pb–Zn deposits to improve exploration. The trace element concentrations of sphalerite can be used to classify the types of Pb–Zn deposits. However, it is difficult to assess the multivariable system through simple data analysis directly. Here, we collected more than 2200 analyses with 14 elements (Mn, Fe, Co, Ni, Cu, Ga, Ge, Ag, Cd, In, Sn, Sb, Pb, and Bi) from 65 deposits, including 48 analyses from carbonate replacement (CR), 684 analyses from distal magmatic-hydrothermal (DMH), 197 analyses from epithermal, 456 analyses from Mississippi Valley-type (MVT), 199 analyses from sedimentary exhalative (SEDEX), 377 analyses from skarn, and 322 analyses from volcanogenic massive sulfide (VMS) types of Pb–Zn deposits. The critical metals in different types of deposits are summarized. Machine learning algorithms, namely, decision tree (DT), K-nearest neighbors (KNN), naive Bayes (NB), random forest (RF), and support vector machine (SVM), are applied to process and explore the classification. Learning curves show that the DT and RF classifiers are the most suitable for classification. Testing of the DT and RF classifier yielded accuracies of 91.2% and 95.4%, respectively. In the DT classifier, the feature importances of trace elements suggest that Ni (0.22), Mn (0.17), Cd (0.13), Co (0.11), and Fe (0.09) are significant for classification. Furthermore, the visual DT graph shows that the Mn contents of sphalerite allow the division of the seven classes into three groups: (1) depleted in Mn, including MVT and CR types; (2) enriched in Mn, including epithermal, skarn, SEDEX, and VMS deposits; and (3) DMH deposits, which have variable Mn contents. Data mining also reveals that VMS and skarn deposits have distinct Co and Ni contents and that SEDEX and DMH deposits have different Ni and Ge contents. The optimal DT and RF classifiers are deployed at Streamlit cloud workspace. Researchers can select DT or RF classifier and input trace element data of sphalerite to classify the Pb–Zn deposit type.https://www.mdpi.com/2075-163X/12/10/1293machine learningsphaleriteLA–ICP–MSPb–Zn depositsweb app |
| spellingShingle | Guo-Tao Sun Jia-Xi Zhou Application of Machine Learning Algorithms to Classification of Pb–Zn Deposit Types Using LA–ICP–MS Data of Sphalerite Minerals machine learning sphalerite LA–ICP–MS Pb–Zn deposits web app |
| title | Application of Machine Learning Algorithms to Classification of Pb–Zn Deposit Types Using LA–ICP–MS Data of Sphalerite |
| title_full | Application of Machine Learning Algorithms to Classification of Pb–Zn Deposit Types Using LA–ICP–MS Data of Sphalerite |
| title_fullStr | Application of Machine Learning Algorithms to Classification of Pb–Zn Deposit Types Using LA–ICP–MS Data of Sphalerite |
| title_full_unstemmed | Application of Machine Learning Algorithms to Classification of Pb–Zn Deposit Types Using LA–ICP–MS Data of Sphalerite |
| title_short | Application of Machine Learning Algorithms to Classification of Pb–Zn Deposit Types Using LA–ICP–MS Data of Sphalerite |
| title_sort | application of machine learning algorithms to classification of pb zn deposit types using la icp ms data of sphalerite |
| topic | machine learning sphalerite LA–ICP–MS Pb–Zn deposits web app |
| url | https://www.mdpi.com/2075-163X/12/10/1293 |
| work_keys_str_mv | AT guotaosun applicationofmachinelearningalgorithmstoclassificationofpbzndeposittypesusinglaicpmsdataofsphalerite AT jiaxizhou applicationofmachinelearningalgorithmstoclassificationofpbzndeposittypesusinglaicpmsdataofsphalerite |