# Mineral Prospecting Prediction via Transfer Learning Based on Geological Big Data: A Case Study of Huayuan, Hunan, China

^{1}

^{2}

^{*}

## Abstract

**:**

## 1. Introduction

## 2. Study Area

#### 2.1. Geological Background

#### 2.1.1. Stratigraphy

#### 2.1.2. Structure

#### Fold Structure

#### Fault Structure

#### 2.1.3. Lithologic Paleogeography

#### 2.2. Mineralization

_{4}+ SO

_{4}

^{2−}

`→`HCO

^{3−}+ HS

^{−}+ H

_{2}O; 2HCO

^{3−}+ Mn

^{2+}

`→`MnCO

_{3}+ CO

_{2}+ H

_{2}O).

_{2}S gas. Meanwhile, H

_{2}S reacts with CO

_{2}in the water body to form HCO

^{3−}, which reacts with manganese ions brought by seawater reflux to form rhombmanganese ore (Figure 3b).

## 3. Research Methods

#### 3.1. Construction of Large Spatial Database

#### 3.2. 3D Modeling and Prediction Layer Construction

#### 3.2.1. Modeling of the Pre-Training Area

#### 3.2.2. Construction of 3D Pre-Training Layer

_{max}= 635,975, X

_{min}= 630,025; Y

_{max}= 3,147,062, Y

_{min}= 3,137,937; Z

_{max}= 450, Z

_{min}= −350).

#### 3.3. 3D CNN Modeling

#### 3.3.1. 3D CNN Algorithm

#### 3.3.2. Localization and Probability Determination

#### 3.4. Transfer Learning Model

#### 3.4.1. Sample Data Expansion

#### 3.4.2. Operating Characteristic Curve

## 4. Results and Discussion

#### 4.1. Comparison of Different Factor Combinations

#### 4.2. Comparative Test on 3DCMM and TL

#### 4.3. Comprehensive Analysis of Prediction Results

## 5. Conclusions

- (1)
- Twenty-two proposed ore-controlling variables were divided into six groups for comparative experiments with different combinations, and each group was further divided into the 3D CNN prediction method and the transfer learning prediction method. After proving the similarity of the regional metallogenic background, the convolution kernel of the Minle area was transferred to that of the Huayuan area with poor data. Then, both were used to train a 3D prediction model to realize the training and transfer of the spatial correlation between the spatial distribution of ore-controlling factors and the manganese ore. The results indicated that the accuracy of the transfer learning model in test 6 could reach 100%, showing the strong stability of the transfer learning prediction model and a fast convergence speed.
- (2)
- By comparing the 3D predicted targets before and after the transfer learning of tests 5 and 6, the 3D CNN model and the prediction models after transfer learning were compared and analyzed in terms of the ROC curve and ore-controlling ratio. It was found that if the 7–12 fracture and each fracture buffer layer are added, the 3D CNN model will perform well, and if they are not added, the transfer learning model will be superior, with an accuracy of 100%. The blocks with high probability values in the prediction results are few and concentrated.
- (3)
- The final prediction results were superimposed and verified with the solid model of the rift basin and the growth fault model in the unpredicted area. The analysis results demonstrated that the delineated metallogenic prospect area had a high degree of overlap with the rift basin, and it is located in the hanging wall of the growth fault, with excellent metallogenic conditions. To sum up, the 3D CNN prediction method has great potential and is advantageous in mineral prediction when big data are available, and transfer learning based on the 3D CNN algorithm greatly improves 3D deep mineral prediction in the case of incomplete data.

## Author Contributions

## Funding

## Conflicts of Interest

## References

- Zhao, P. Digital mineral exploration and quantitative evaluation in the big data. Geol. Bull. China
**2015**, 34, 1255–1259. [Google Scholar] - Lohr, S. The Age of Big Data. The New York Times, 11 February 2012; 1–5. [Google Scholar]
- Bristol, R.S.; Euliss, N.H., Jr.; Booth, N.L.; Burkardt, N.; Diffendorfer, J.E.; Gesch, D.B.; McCallum, B.E.; Miller, D.M.; Morman, S.A.; Poore, B.S.; et al. Science Strategy for Core Science Systems in the U.S. Geological Survey 2013–2023: Public Review Release; U.S. Geological Survey: Reston, VA, USA, 2012. [Google Scholar]
- Chen, S. Research of Multiple Geoscience Information Prospecting Prediction in Xikuangshan Antimony Ore Field; China University of Geosciences: Beijing, China, 2012. [Google Scholar]
- Zhang, D. Spatially Weighted Technology for Logistic Regression and Its Application in Mineral Prospectivity Mapping; China University of Geosciences: Wuhan, China, 2015. [Google Scholar]
- Lin, N. Study on the Metallogenic Prediction Models Based on Remote Sensing Geology and Geochemical Information Case Study of Lalingzaohuo Region in Qinghai Province; Jilin University: Changchun, China, 2015. [Google Scholar]
- Yan, G.; Xue, Q.; Xiao, K.; Chen, J.; Miao, J.; Yu, H. An analysis of major problems in geological survey big data. Geol. Bull. China
**2015**, 34, 1273–1279. [Google Scholar] - Zhang, Q.; Zhou, Y. Reflections on the scientific research method in the era of big data. Bull. Mineral. Petrol. Geochem.
**2017**, 52, 637–648. [Google Scholar] - Zhou, Y.; Li, P.; Wang, S.; Xiao, F.; Li, J.; Gao, L. Research Progress on Big Data and Intelligent Modelling of Mineral Deposits. Bull. Mineral. Petrol. Geochem.
**2017**, 36, 327–331. [Google Scholar] - Zhou, Y.Z.; Chen, S.; Zhang, Q.; Xiao, F.; Wang, S.G.; Liu, Y.P.; Jiao, S.T. Advances and prospects of big data and mathematical geoscience. Acta Petrol. Sin.
**2018**, 34, 255–263. [Google Scholar] - LeCun, Y.; Bottou, L.; Bengio, Y.; Haffner, P. Gradient-based learning applied to document recognition. Proc. IEEE
**1998**, 86, 2278–2324. [Google Scholar] [CrossRef] - Lawrence, S.; Giles, C.L.; Tsoi, A.C.; Back, A.D. Face recognition: A convolutional neural-network approach. IEEE Trans. Neural Netw.
**1997**, 8, 98–113. [Google Scholar] [CrossRef] - Simonyan, K.; Zisserman, A. Very Deep Convolutional Networks for Large-Scale Image Recognition. arXiv
**2014**, arXiv:1409.1556. [Google Scholar] - Sun, S.; He, Y. Multi-label emotion classification for microblog based on CNN feature space. Adv. Eng. Sci.
**2017**, 49, 162–169. [Google Scholar] - Martens, J.; Sutskever, I. Learning recurrent neural networks with hessian-free optimization. In Proceedings of the 28th International Conference on Machine Learning, Washington, DC, USA, 28 June–2 July 2011; ACM Press: New York, NY, USA, 2011; pp. 1033–1040. [Google Scholar]
- Bengio, Y.; Lamblin, P.; Popovici, D.; Larochelle, H. Greedy Layer-Wise Training of Deep Networks. In Advances in Neural Information Processing Systems 19, Proceedings of the Twentieth Annual Conference on Neural Information Processing Systems, Vancouver, BC, Canada, 4–7 December 2006; Schölkopf, B., Platt, J., Hofmann, T., Eds.; The MIT Press: Cambridge, MA, USA, 2007; pp. 153–160. [Google Scholar]
- Sainath, T.N.; Kingsbury, B.; Ramabhadran, B. Auto-encoder bottleneck features using deep belief networks. In Proceedings of the 2012 International Conference on Acoustics, Speech and Signal Processing, Kyoto, Japan, 25–30 March 2012; pp. 4153–4156. [Google Scholar]
- Hinton, G. Deep belief nets. In Encyclopedia of Machine Learning; Springer: Boston, MA, USA, 2011; pp. 267–269. [Google Scholar]
- Ciresan, D.; Giusti, A.; Gambardella, L.M.; Schmidhuber, J. Deep neural networks segment neuronal membranes in electron microscopy images. In Proceedings of the 25th International Conference on Neural Information Processing Systems, Granada, Spain, 12–15 December 2011; Curran Associates Inc.: Red Hook, NY, USA, 2012; pp. 2843–2851. [Google Scholar]
- Holtzman, B.K.; Paté, A.; Paisley, J.; Waldhauser, F.; Repetto, D. Machine learning reveals cyclic changes in seismic source spectra in Geysers geothermal field. Sci. Adv.
**2018**, 4, eaao2929. [Google Scholar] [CrossRef] - Rouet-Leduc, B.; Hulbert, C.; Lubbers, N.; Barros, K.; Humphreys, C.J.; Johnson, P.A. Machine learning predicts laboratory earthquakes. Geophys. Res. Lett.
**2017**, 44, 9276–9282. [Google Scholar] [CrossRef] - Zhou, Y.Z.; Wang, J.; Zuo, R.G.; Xiao, F.; Shen, W.J.; Wang, S.G. Machine learning, deep learning and Python language in field of geology. Acta Petrol. Sin.
**2018**, 34, 3173–3178. [Google Scholar] - Zhang, Y.; Li, M.C.; Han, S. Automatic identification and classification in lithology based on deep learning in rock images. Acta Petrol. Sin.
**2018**, 34, 333–342. [Google Scholar] - Zhou, Q.; Du, Y.; Yuan, L.; Zhang, S.; Yang, B.; Pan, W.; Wang, P.; Yu, W.; Xu, Y.; Qi, L.; et al. Exploration models of ancient natural gas seep sedimentary-type manganese ore deposit: A case study of the Nanhua period Datangpo type manganese ore in the conjunction area of Guizhou, Hunan and Chongqing. Acta Geol. Sin.
**2017**, 91, 2285–2298. [Google Scholar] - Zhou, Q.; Du, Y.; Qin, Y. Ancient natural gas seepage sedimentary-type manganese metallogenic system and ore-forming model: A case study of Datangpo type manganese deposits for med in rift basin of Nanhua Period along Guizhou-Hunan-Chongqing border area. Miner. Depos.
**2013**, 32, 457–466. [Google Scholar] - Du, Y.S.; Zhou, Q.; Yu, W.C.; Wang, P.; Yuan, L.; Qi, L.; Guo, H.; Xu, Y. Linking the Cryogenian Manganese Metallogenic Process in the Southeast Margin of Yangtze Block to Break-Up of Rodinia Supercontinent and Sturtian Glaciation. Geol. Sci. Technol. Inf.
**2015**, 34, 1–7. [Google Scholar] - Yang, L. Geology and Geochemistry Characters Study of Guzhang Lannitian Manganese Deposit in West Hunan Province; Central South University: Changsha, China, 2014. [Google Scholar]
- Zou, G.-J.; Huang, J.-Z.; Ling, Y.-X.; Tan, S.M.; Cao, C.H.; Huang, G.F.; Wen, C.H. Lithofacies Palaeogeography and Sedimentary Evolution of Hunan–Guizhou–Chongqing Border Area in the Nanhua Period. Acta Geosci. Sin.
**2020**, 41, 207–218. [Google Scholar] - Li, S.; Chen, J.; Liu, C.; Wang, Y. Mineral Prospectivity Prediction via Convolutional Neural Networks Based on Geological Big Data. J. Earth Sci.
**2021**, 32, 21. [Google Scholar] [CrossRef] - Chen, H.; Qi, X.; Yu, L.; Heng, P.-A. DCAN: Deep contour-aware networks for accurate gland segmentation. In Proceedings of the Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 27–30 June 2016; pp. 2487–2496. [Google Scholar]
- Kamnitsas, K.; Ledig, C.; Newcombe, V.F.J.; Simpson, J.P.; Kane, A.D.; Menon, D.K.; Rueckert, D.; Glocker, B. Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation. Med. Image Anal.
**2017**, 36, 61–78. [Google Scholar] [CrossRef] - Simard, P.Y.; Steinkraus, D.; Platt, J.C. Best practices for convolutional neural networks applied to visual document analysis. In Proceedings of the Seventh International Conference on Document Analysis and Recognition, Edinburgh, UK, 6 August 2003; Volume 2, pp. 958–962. [Google Scholar]
- Dieleman, S.; Willett, K.W.; Dambre, J. Rotation-invariant convolutional neural networks for galaxy morphology prediction. Mon. Not. R. Astron. Soc.
**2015**, 450, 1441–1459. [Google Scholar] [CrossRef] - Krizhevsky, A.; Sutskever, I.; Hinton, G.E. Image Net classification with deep convolutional neural networks. In Proceedings of the 25th International Conference on Neural Information Processing Systems, Lake Tahoe, NV, USA, 3–6 December 2012; Curran Associates Inc.: Red Hook, NY, USA, 2012; pp. 1097–1105. [Google Scholar]
- Hinton, G.; Deng, L.; Yu, D.; Dahl, G.E.; Mohamed, A.R.; Jaitly, N.; Senior, A.; Vanhoucke, V.; Nguyen, P.; Sainath, T.N.; et al. Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups. IEEE Signal Process. Mag.
**2012**, 29, 82–97. [Google Scholar] [CrossRef] - Tong, L.; Renguang, Z.; Yihui, X.; Peng, Y. Random-Drop Data Augmentation of Deep Convolutional Neural Network for Mineral Prospectivity Mapping. Nat. Resour. Res.
**2020**, 30, 27–38. [Google Scholar] - Chen, Y. Mineral potential mapping with a restricted Boltzmann machine. Ore Geol. Rev.
**2015**, 71, 749–760. [Google Scholar] [CrossRef] - Chen, Y.; Wu, W. A prospecting cost−benefit strategy for mineral potential mapping based on ROC curve analysis. Ore Geol. Rev.
**2016**, 74, 26–38. [Google Scholar] [CrossRef] - Gao, Y.; Zhang, Z.; Xiong, Y.; Zuo, R. Mapping mineral prospectivity for Cu polymetallic mineralization in southwest Fujian Province, China. Ore Geol. Rev.
**2016**, 75, 16–28. [Google Scholar] [CrossRef] - Parsa, M.; Maghsoudi, A.; Yousefi, M. A receiver operating characteristics−Based geochemical data fusion technique for targeting undiscovered mineral deposits. Nat. Resour. Res.
**2018**, 27, 15–28. [Google Scholar] [CrossRef] - Wang, J. Identification of Geochemical Anomalies Based on Geostatistical Simulation; China University of Geosciences: Wuhan, China, 2018. [Google Scholar]
- Fawcett, T. An introduction to ROC analysis. Pattern Recognit. Lett.
**2006**, 27, 861–874. [Google Scholar] [CrossRef] - Bradley, A.P. The use of the area under the ROC curve in the evaluation of machine learning algorithms. Pattern Recognit.
**1997**, 30, 1145–1159. [Google Scholar] [CrossRef] - Hanley, J.A.; McNeil, B.J. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology
**1982**, 143, 29–36. [Google Scholar] [CrossRef] - Agterberg, F.P.; Bonham-Carter, G.F. Measuring the Performance of Mineral-Potential Maps. Nat. Resour. Res.
**2005**, 14, 1–17. [Google Scholar] [CrossRef]

**Figure 1.**The regional geological and mineral map and district plan at a scale of 1:200,000. (

**A**) represents the Songtao–Huayuan area; (

**B**) represents the Huayuan Mining area; (

**C**) represents the transfer learning pre-training area (Minle); and (

**D**) represents the area with detailed data (according to the Hunan Bureau of Geology and Mineral Exploration and Development).

**Figure 2.**(

**a**) The lithofacies paleogeography map of Songtao-Huayuan in the Nanhua period and Datangpo period. (

**b**) The sequence stratigraphic division map of the Nanhua period (provided by the Hunan Geological Survey Institute).

**Figure 3.**(

**a**) The metallogenic model of the regional manganese ore. (

**b**) The evolutionary model of the regional sedimentary manganese basin [28].

**Figure 6.**The structure of the 3D CNN prediction model. The first layer is the input layer with 22 voxels, and the size of the voxels is 100 m × 100 m × 100 m. Conv3D1 and Conv3D2 are convolutional layers, and the filter size of both layers is 5 × 5 × 5. MaxPooling3d1 and MaxPooling3d2 are 3D max-pooling layers with a pooling sizes of 2 × 2 × 2. Dropout is applied between the flatten layer and the dense layer, and the fraction of the input units to drop is set to 0.5. Note that the regularization method is applied on the dense layer.

**Figure 8.**Classification of transfer learning (

**a**) based on characteristic space and label space and (

**b**) based on the transfer learning method.

**Figure 9.**The principle of sample expansion [36].

**Figure 10.**The spatial relationship and the confusion matrix: (

**a**) the spatial relationship between prediction results and known mineral deposits (points) and (

**b**) the confusion matric for four types of relationships [41].

**Figure 12.**The output of the convolution kernel under the 3D CNN transfer learning in the Minle mining area.

**Figure 13.**Comparison of the confusion matrix and ROC curves of the six groups of transfer learning models.

**Figure 14.**The accuracy and loss curves of the six groups of transfer learning models. (

**a**–

**f**) respectively shows the training loss, training accuracy, validation accuracy and validation loss of the transfer learning model of the experimental 1–6.

**Figure 15.**The performance evaluation curves of mineral prediction by the 3D CNN model and the 3D CNN-TL model in six tests. (

**a**,

**d**) show the success rate curve and ore-controlling rate curve of twelve tests, respectively. (

**b**) shows the success rate curve of the 3D CNN model in six tests. (

**c**) shows the success rate curve of the 3D CNN-TL model in six tests. (

**e**,

**f**) were plotted by extracting the ore-controlling rate between 0.95 and 1 of the curve in (

**d**). (

**e**) shows the ore-controlling rate curve of the 3D CNN model in six tests. (

**f**) shows the ore-controlling rate curve of the 3D CNN-TL model in six tests.

**Figure 16.**The results comparison of 3D metallogenic prospect delineation in test 5. (

**a**,

**c**,

**e**) show the prediction results of the 3DCNN-TL model; (

**b**,

**d**,

**f**) show the prediction results of the 3D CNN. (

**g**) shows the target section delineated by the 3D CNN-TL prediction model when the threshold value of the metallogenic prospect is 0.65, and (

**h**) shows the target section delineated by the 3D CNN model.

**Figure 17.**The result comparison of 3D metallogenic prospect delineation in test 6. (

**a**,

**c**,

**e**) show the prediction results of the 3D CNN-TL model; (

**b**,

**d**,

**f**) show the prediction results of the 3D CNN. (

**g**) is the target section delineated by the 3D CNN-TL prediction model when the threshold value of the metallogenic prospect is 0.65, and (

**h**) is the target section delineated by the 3D CNN model.

**Figure 18.**The overlayed diagrams of 3D predicted results and basin and growth faults: (

**a**) the 3D model of inferred extensional basins, (

**b**) the 3D overlayed diagram of inferred extensional basins and growth faults, (

**c**) the overlayed diagram of prediction results and basins and faults under the 3D CNN model in test 5, and (

**d**) the overlayed diagram of transfer learning in test 6 with basins and faults (the threshold value is 0.65).

Ore Deposit Type | Factor Type | Ore-Controlling Factors |
---|---|---|

“Datangpo type” sedimentary manganese ore | Rock | Rock strata: interglacial period, thick moraine conglomerate |

Morphology of ore body | Lenticular and interlaminar | |

Stratigraphic marks | Metallogenic age: Datangpo formation of the Nanhua period | |

Manganese-bearing rock series outcropped | ||

Speculated distribution of manganese-bearing rock series | ||

Structural marks | Manganese-forming sedimentary basin | |

Syndepositional fault | ||

Lithofacies paleogeography | ||

Geophysics | Gravity anomalies | |

Gravity anomaly transition zone | ||

Geochemistry | Mn geochemical anomaly | |

P geochemical anomaly | ||

Y geochemical anomaly |

Data Name | Scale | Number | Investigation Depth (m) |
---|---|---|---|

Statistical table of geochemical element content | 1:200,000 | 39 | Earth’s surface |

Geochemistry mapping | 1:200,000 | 20 | Earth’s surface |

Geological and mineral map of the Songtao-Huayuan area | 1:200,000 | 1 | Earth’s surface |

Lithofacies paleogeography map of the Datangpo period in the Songtao-Huayuan area | 1:100,000 | 1 | Earth’s surface |

Distribution map of the Datangpo Formation in the Songtao-Huayuan area | 1:100,000 | 1 | Earth’s surface |

Sedimentary tectonic map of the Songtao-Huayuan area | 1:100,000 | 1 | Earth surface |

Residual gravity anomaly map in the Songtao-Huayuan area | 1:100,000 | 1 | - |

MT profile inversion map | 1:100,000 | 2 | 4000 |

Data Name | Scale | Number | Investigation Depth (m) |
---|---|---|---|

Prospecting line profile map of Minle | 1:2000 | 46 | 700 |

Geological map of Heku and Malichang in the Huayuan area | 1:50,000 | 1 | Earth’s surface |

Borehole data of the Minle Mn mine | - | 212 | 660 |

Lithofacies paleogeography of the Datangpo period in the Songtao-Huayuan area | 1:100,000 | 1 | Earth’s surface |

Schematic diagram of the cut section design | 1:50,000 | 1 | Earth’s surface |

Map cut profiles of the Minle Mn mine | 1:50,000 | 23 | 1500 |

DEM data | Aster30m | 1 | Earth’s surface |

Data Name | Scale | Number | Investigation Depth (m) |
---|---|---|---|

Borehole data of the Minle Mn mine | - | 212 | 660 |

Prospecting line profile map of Minle | 1:2000 | 46 | 700 |

Lithofacies paleogeography of the Datangpo period in the Songtao-Huayuan area | 1:100,000 | 1 | Earth’s surface |

DEM data | Aster30m | 1 | Earth’s surface |

Topographic geological map of the Minle Mn mining area | 1:10,000 | 1 | Earth’s surface |

Schematic diagram of the cut section design | 1:10,000 | 1 | Earth’s surface |

Map cut profiles of the Minle Mn mine | 1:10,000 | 90 | 700 |

Ore-Controlling Factors | Metallogenic Geological Anomaly | Ore Prospecting Digital Model |
---|---|---|

Stratum | The stratum of the Datangpo Formation | Stratum digital model of the Datangpo Formation |

Rock stratum indicator | The thick moraine conglomerate, as the top plate of the ore bed, can indicate the mineralization | Digital model of the Nantuo moraine layer |

Structure | Deep fracture and tectonic convergence (submarine volcanic eruption, sedimentary manganese deposit, and ancient natural gas leakage) | Fracture digital model |

Lithofacies paleogeography | The secondary rift basin | Digital model of the basin control fracture and basin margin fracture |

Ore body | The manganese ore body | 3D digital model of the manganese ore body |

Ore-Controlling Factors | Metallogenic Geological Anomaly | 3D Spatial Reconstruction of Metallogenic Anomaly |
---|---|---|

Stratum | Sub-members under the Datangpo Formation | 3D spatial reconstruction of the lower sub-members under the Datangpo Formation |

3D spatial reconstruction of the medium sub-members under the Datangpo Formation | ||

3D spatial reconstruction of the upper sub-members under the Datangpo Formation | ||

Sequence stratigraphy | Favorable third-sequence stratigraphy | 3D spatial reconstruction of the condensation layer |

3D spatial reconstruction of the high-stand domain | ||

Rock stratum indicator | Favorable rock stratum indicator | 3D spatial reconstruction of manganese-bearing rock series |

Lithofacies paleogeography | Favorable sedimentary facies | Secondary rift basin |

Ancient landform | Ancient landform of the geological evolution period related to mineralization | 3D spatial reconstruction of the ancient landform |

Predictor Numbers | Evidence Item | W+ | W- | C |
---|---|---|---|---|

Predictor 1 | The third sequence | 4.337955 | 1.802756 | 2.5352 |

Predictor 2 | Datangpo Formation | 3.620267 | 1.877229 | 1.743037 |

Predictor 3 | Favorable metallogenic buffer zone of the Nantuo moraine layer | 4.367813 | −0.36435 | 4.732164 |

Predictor 4 | Interlaminar anomalous body of ancient landform of Datangpo Formation | 3.546471 | −7.98981 | 11.53628 |

Predictor 5 | Interlaminar anomalous body of ancient landform of Nantuo Formation | 2.821807 | 2.024608 | 0.797199 |

Predictor 6 | Interlaminar anomalous body of ancient landform of Nanhua Period | 2.866911 | 1.833689 | 1.033222 |

Predictor 7 | Fault | 2.861902 | 1.564909 | 1.296994 |

Predictor 8 | 50 m Fault buffer | 2.776001 | 1.329269 | 1.446732 |

Predictor 9 | 100 m Fault buffer | 2.673975 | 1.17519 | 1.498785 |

Predictor 10 | 150 m Fault buffer | 2.535793 | 0.878388 | 1.657405 |

Predictor 11 | 200 m Fault buffer | 0.687016 | 2.197602 | −1.51059 |

Predictor 12 | 250 m Fault buffer | 2.535793 | 0.878388 | 1.657405 |

Predictor 13 | Interlaminar anomalous body of ancient landform of Fulu Formation | 2.535793 | 0.878388 | 1.657405 |

Predictor 14 | 300 m Fault buffer | 2.264051 | 2.09229 | 0.171761 |

Predictor 15 | 350 m Fault buffer | 0 | 0 | 0 |

Predictor 16 | Equidensity | 2.540403 | 2.085783 | 0.45462 |

Predictor 17 | Azimuth anomalous degree | 2.624474 | 2.092335 | 0.532139 |

Predictor 18 | Anomalous azimuth | 2.2477 | 2.091067 | 0.156634 |

Predictor 19 | Centrosymmetry degree | 0 | 0 | 0 |

Predictor 20 | Normalization frequency | 0 | 0 | 0 |

Predictor 21 | Number of normalization intersections | 4.337955 | 1.802756 | 2.5352 |

Predictor 22 | Nantuo moraine layer | 3.620267 | 1.877229 | 1.743037 |

No | Prediction Factors/pcs | 3D Prediction Layers Combination | Network Size/m | Iteration Frequency/Times | Training Accuracy | Training Loss | Validation Accuracy /% | Validation Loss |
---|---|---|---|---|---|---|---|---|

1 | 22 | 1–22 | 100 × 100 × 100 | 215 | 100% | 0.000021 | 100% | 0.000116 |

2 | 21 | 1–21 (remove 22) | 100 × 100 × 100 | 215 | 100% | 0.000203 | 100% | 0.000299 |

3 | 15 | 1–15 (remove 16–22) | 100 × 100 × 100 | 215 | 100% | 0.000576 | 100% | 0.000741 |

4 | 13 | 1–13 (remove 14–22) | 100 × 100 × 100 | 215 | 100% | 0.000004 | 100% | 0.000099 |

5 | 12 | 1–12 (remove 13–22) | 100 × 100 × 100 | 215 | 100% | 0.000105 | 100% | 0.000079 |

6 | 6 | 1–6 (remove 7–22) | 100 × 100 × 100 | 215 | 100% | 0.000022 | 100% | 0.000022 |

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |

© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Li, S.; Liu, C.; Chen, J.
Mineral Prospecting Prediction via Transfer Learning Based on Geological Big Data: A Case Study of Huayuan, Hunan, China. *Minerals* **2023**, *13*, 504.
https://doi.org/10.3390/min13040504

**AMA Style**

Li S, Liu C, Chen J.
Mineral Prospecting Prediction via Transfer Learning Based on Geological Big Data: A Case Study of Huayuan, Hunan, China. *Minerals*. 2023; 13(4):504.
https://doi.org/10.3390/min13040504

**Chicago/Turabian Style**

Li, Shi, Chang Liu, and Jianping Chen.
2023. "Mineral Prospecting Prediction via Transfer Learning Based on Geological Big Data: A Case Study of Huayuan, Hunan, China" *Minerals* 13, no. 4: 504.
https://doi.org/10.3390/min13040504