Machine learning based multivariate data prospectivity system

Abstract

The machine learning-based multi-element data prospecting prediction system comprises a module 1 deposit model construction module, a module 2 spatial database construction module, a module 3 data standardization module, a module 4 prediction model training data distribution module, a module 5 machine learning-based prospecting prediction model construction module, a module 6 algorithm optimization and parameter adjustment module and a module 7 potential area analysis and target area delineation module, and can sequentially perform the following steps: preferential selection of a favorable ore-forming layer, comprehensive processing of multi-element information and establishment of a multi-element spatial database, data standardization and gridding of the database, multi-model multi-parameter ore-forming potential analysis based on artificial neural networks and support vector machines, quantitative evaluation of the prediction model by using an ROC receiver operating characteristic curve and determination of an optimal prospecting prediction model, and determination of a threshold value by using a Youden index and delineation of an ore-forming prospective prediction area or target area distribution map.

Description

Technical Field

[0001] This invention relates to a cutting-edge technology field that intersects multiple disciplines, including geology, geophysics, geochemistry, remote sensing technology, artificial intelligence, mineral deposit geology, geographic information systems, and computer technology. Specifically, it relates to a multi-data mineral exploration prediction system based on machine learning. Background Technology

[0002] Mineral exploration prediction is a highly comprehensive task. Its working principle involves studying existing geological, mineral, geophysical, geochemical, and remote sensing data of the target area to deeply analyze the metallogenic characteristics, geological laws, and the spatiotemporal distribution, size, type, surrounding rocks, intrusive rock influence, and tectonic damage relationships of known mineral deposits. Based on the corresponding metallogenic geological laws, metallogenic geological conditions, ore-controlling factors, mineral exploration indicators, and mineral exploration prediction theories, metallogenic models and mineral exploration prediction models are summarized to delineate and evaluate prospective areas within the study area and guide the next stage of geological exploration.

[0003] As geological work becomes more sophisticated, mineral exploration faces increasing challenges, and traditional approaches and methods are no longer sufficient for breakthroughs. Furthermore, with the widespread application of advanced technologies in mineral exploration, geological engineers are obtaining increasingly richer data through various exploration techniques, making prediction more difficult. Therefore, leveraging machine learning methods to harness data-driven innovation and analyze, mine, and process this increasingly diverse data to solve the growing challenges of mineral prediction is of great significance.

[0004] Our research group has not found any previously published literature of the same type.

[0005] Therefore, it is of great significance and urgent to develop a multi-data mineral exploration prediction system based on machine learning that can analyze, mine and process multi-data. Summary of the Invention

[0006] The objective of this invention is to overcome the shortcomings of existing technologies and provide a multi-data mineral exploration prediction system based on machine learning, which can analyze, mine and process multi-data and make scientific predictions for mineral exploration.

[0007] The objective of this invention is achieved through the following technical solution:

[0008] A machine learning-based multi-data mineral exploration prediction system addresses the challenges of mineral exploration by comprising: Module 1, Mineral Deposit Model Construction; Module 2, Spatial Database Construction; Module 3, Data Standardization; Module 4, Predictive Model Training Data Allocation; Module 5, Machine Learning-Based Mineral Exploration Prediction Model Construction; Module 6, Algorithm Optimization and Parameter Tuning; and Module 7, Potential Area Analysis and Target Area Delineation. This system sequentially performs the following: selection of favorable mineralization layers; comprehensive processing of multi-source information and establishment of a multi-source spatial database; data standardization and gridding of the database; multi-model, multi-parameter mineralization potential analysis based on machine learning artificial neural networks and support vector machines; quantitative evaluation of the prediction model using ROC (Receiving Characteristic Curve) and determination of the optimal mineral exploration prediction model; and determination of thresholds and delineation of mineralization prospective prediction areas or target area distribution maps using the Youden index.

[0009] Compared with the prior art, the present invention has the following advantages or effects:

[0010] (1) It has outstanding advantages in processing complex geological data. Machine learning algorithms can process high-latitude and nonlinear multi-data, and quantify traditional qualitative prediction methods, which is a significant improvement.

[0011] (2) It can improve the accuracy of prediction. By training and learning favorable information for mineral exploration such as geology, geophysics, chemistry and remote sensing, the machine learning model can explore the logistic regression relationship between each input layer and the known mineral deposits and mineralization points, thereby improving the accuracy of prediction. Compared with prediction methods that rely entirely on expert experience, the method based on the combination of data-driven machine learning and expert knowledge is more objective and can reduce the deviation that may be caused by human subjective judgment.

[0012] (3) It can combine multi-dimensional geoscience data, such as geology, geochemistry, geophysics, etc., to make comprehensive mineralization predictions, which helps to reduce the multiple interpretations of single information and the uncertainty of mineral exploration. It can reduce the uncertainty of mineral exploration prediction by optimizing feature selection and dataset composition, improve the generalization ability of the model, and can be extended to different geological environments and mineral deposit types, especially regional target area optimization, with good prediction effect.

[0013] (4) The degree of automation and efficiency have been significantly improved. After the mineral exploration prediction model is trained, it can quickly predict new data, which greatly improves the efficiency of mineral exploration work.

[0014] (5) It can effectively extract characteristic factors from a large amount of raw geological data, perform dimensionality reduction processing, make the results easier to understand, generate intuitive prediction result maps, and facilitate geological experts to conduct further analysis and decision-making.

[0015] (6) It has the ability to continuously learn and optimize. As time goes by and data accumulates, the machine learning mineral exploration prediction model can continuously optimize its prediction ability through continuous learning. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the module structure of a multi-data mineral exploration prediction system based on machine learning proposed in this invention.

[0017] Figure 2 for Figure 1 The diagram shows the working principle and process flow of the module.

[0018] The present invention will now be described in further detail with reference to the accompanying drawings. Detailed Implementation

[0019] like Figures 1-2 As shown, to address the challenges of mineral exploration, the module includes Module 1 (Mineral Deposit Model Construction), Module 2 (Spatial Database Construction), Module 3 (Data Standardization), Module 4 (Predictive Model Training Data Allocation), Module 5 (Machine Learning-Based Mineral Exploration Prediction Model Construction), Module 6 (Algorithm Optimization and Parameter Tuning), and Module 7 (Potential Area Analysis and Target Area Delineation). These modules sequentially perform the following: selection of favorable mineralization layers; comprehensive processing of multi-source information and establishment of a multi-source spatial database; data standardization and gridding of the database; multi-model, multi-parameter mineralization potential analysis based on machine learning artificial neural networks and support vector machines; quantitative evaluation of the prediction model using ROC (Receiving Characteristic Curve) and determination of the optimal mineral exploration prediction model; and determination of thresholds and delineation of mineralization prospective prediction areas or target area distribution maps using the Youden index.

[0020] The process of the present invention can be further described as follows:

[0021] Module 1 is used to analyze the geological characteristics of mineralization and the location patterns of ore bodies; Module 2 is used to unify the scope, coordinates, and data format of the analysis data; Module 3 is used to solve the sample imbalance of the analysis data; Module 4 is used to distribute the input sample data into two categories: training data and test data; Module 5 is used to construct a mineral exploration prediction model; Module 6 is used to optimize the algorithm and adjust the parameters of the constructed mineral exploration prediction model to achieve the optimal solution; Module 7 is used to use the optimal mineral exploration prediction model to analyze potential areas and complete the target area delineation through interaction with experts.

[0022] The working steps and conditions of the module are as follows: (1) Collect and sort out the geological data such as the regional metallogenic geological conditions, strata, structure, and magmatic activity in the study area, and analyze the geological characteristics of typical mineral deposits in the area; (2) Sort out and statistically analyze the physical parameters in the study area, perform secondary processing and anomaly information extraction on the collected magnetic data, compare the relationship between different anomaly units and known mineral deposits, and determine the geophysical elements related to mineralization; (3) Analyze and summarize the regional geochemical characteristics and the geochemical characteristics of the study area, determine the lower limit of anomalies using the iterative method and the Youden index method respectively, and delineate single-element anomaly maps; use R-type cluster analysis to classify the elements, delineate two types of combined anomaly maps, and extract geochemical elements related to mineralization; (4) Collect image data and DEM data of the study area, and use principal component analysis to analyze the image data and DEM data of the study area. The analysis method extracted mineralization, alteration information and linear structure; (5) extracted geological elements related to mineralization, and established a multi-dimensional mineralization information spatial database for elements favorable to mineralization, and standardized and gridded the data; (6) used machine learning models such as multilayer perceptron (MLP) and support vector machine (SVM) to conduct multi-parameter mineralization potential analysis and establish multiple mineral exploration prediction models; (7) used ROC curve analysis technology and AUC index to quantitatively evaluate multiple mineral exploration prediction models and determine the optimal mineral exploration prediction model; (8) used Youden index to delineate the mineral exploration prediction area or mineral exploration target area for the data of the optimal mineral exploration prediction model, and used GIS visualization spatial analysis technology and expert knowledge-driven qualitative evaluation of the mineral exploration prediction model indicators of typical deposits in the study area.

[0023] The step (1) of systematically collecting exploration technology results data in the study area specifically includes: geological data, geophysical data, geochemical data, and remote sensing data.

[0024] The geological data in step (1) includes regional metallogenic belts, stratigraphic units, igneous rocks, tectonic-magmatic environments, metallogenic epochs, ore deposits, mineralization points, metallogenic geological bodies, veins, metallogenic structures and structural planes, ore body characteristics of typical ore deposits, mineral assemblages, ore structure and texture, mineralization and alteration types, etc.

[0025] The geophysical data in step (2) includes electrical data, magnetic data, and gravity data.

[0026] The geochemical data in step (3) includes river sediment measurement data, soil geochemical measurement data, and rock geochemical measurement data.

[0027] The remote sensing data in step (4) includes high-resolution image data, DEM data, multispectral data, etc.

[0028] The step (5) of selecting favorable mineralization layers specifically involves: mineral deposits have specific mineralization patterns, mineralization characteristics, mineralization backgrounds, mineralization conditions, and other mineralization indicators for mineralization prediction. All geophysical, geochemical, and remote sensing information is associated with mineral deposits and mineralization points. Factor layers favorable to mineralization are extracted, and all mineralization favorable factor information is processed by machine learning to achieve mineralization prediction and breakthroughs.

[0029] Step (5) involves extracting favorable geochemical mineralization factors based on the spatial distribution patterns of high, low, and high background elements in the region, including copper-affinity elements Cu, Au, Ag, Pb, Zn, Cd, As, Sb, Hg; tungsten-molybdenum group elements W, Mo, Sn, Bi, B, F; iron group elements Fe, Cr, Ni, Co, V, Ti, Mn, Mg; and rare, rare earth, and radioactive elements Li, Be, Nb, Zr, La, Y, Th, U. The Youden index method is innovatively used to calculate the lower limit of anomalies in the original geochemical data. This is compared with the iterative estimation method. Analysis of the lower limit values ​​obtained by the two algorithms reveals that the Youden index method yields a higher lower limit value and fewer invalid anomalies. A better lower limit threshold is adopted. GIS software is used to interpolate and generate contour maps of each element. R-type cluster analysis is then performed to statistically identify element combinations related to the deposit type and generate an element combination anomaly map.

[0030] The extraction of favorable factors for mineralization by remote sensing in step (5) mainly involves linear structural extraction and alteration information extraction. The remote sensing data used are Landsat 8 satellite OLI 15m data, Landsat 8 satellite OLI 30m data, and ASTGTM2_dem data. The alteration information is extracted directly by principal component analysis using a 30-meter resolution multispectral image. Based on the multispectral data, the information on iron staining anomaly and mud mineralization alteration is extracted. The OLI564 color composite image is formed by fusing the bands of the 30-meter and 15-meter resolution images for remote sensing geological interpretation. Structural interpretation lines are extracted. The linear structural extraction uses three-dimensional scene construction and stereoscopic visualization extraction technology. The image fusion platform is ENVI5.2. The NNDiffuse Pan Sharpeniing algorithm is used for image fusion to ensure that the fused image retains good color texture and spectral information. The resolution of the fused image reaches 15 meters. The spectral information includes seven bands, such as shortwave infrared, near-infrared, and visible light, which are used for geological research.

[0031] The step (5) involves the comprehensive processing of multi-source information to establish a multi-source spatial database. Specifically, this involves cropping, registering, correcting, and converting all layers of favorable mineralization factors in the mineralization model and prospecting indicators to ensure that all layers have the same range, coordinates, and data format, and ultimately achieve automatic overlay and fitting of layer information.

[0032] Step (5) database data standardization and gridding specifically involves: dividing multi-dimensional information according to the gridded block size to form unified grid block units. The gridded unit size depends on the scale, with the default block size being 50*50m. Deviation will affect the training speed of machine learning. The grid block units are assigned attribute values ​​based on whether the multi-dimensional mineralization information is covered. The mineralization information of the grid block units is distinguished by binary assignment. The location covered by mineralization information is assigned a value of 1, and the blank area is assigned a value of 0. The relationships between all layers extracted from geology, geophysics, geology, and remote sensing, as well as between layers and mineralization patterns and prospecting indicators, are comprehensively analyzed to determine the correlation between the standardized data and mineralization. Other redundant layers are removed. For linear structural layers that cannot be gridded, areal buffer and structural density map analysis techniques are used. The areal buffer analysis technique involves selecting buffers of different radii. A comparative study was conducted within the study area, with a default step size of 100m. The ratio of the number of mineral points to the area within the buffer zone was used as the analysis coefficient. By analyzing different buffer zone coefficients, the spatial range most significantly controlled by linear structures was determined. The structural density map analysis technique determined the grid cell size by analyzing the length and pattern of linear structures within the study area. The study area was divided into grid cells, with a default size of 100*100m. All linear structures and grid cells within the study area were overlaid and topologically reconstructed, so that each linear structure was clipped by the grid cell. The lengths of all clipped linear structures within each grid cell were summed, and the total length of linear structures within each grid cell was calculated. The structural density coefficient of all grid cells was calculated by dividing the total length of linear structures by the area of ​​each grid cell. The structural density coefficient attribute of each grid cell was extracted into points, and raster interpolation was used to generate a structural density contour map of the study area.

[0033] The analysis of mineralization potential using machine learning-based artificial neural networks and support vector machines (SVML) involves multiple models and parameters, specifically considering the specific characteristics of mineralization scale and prospecting indicators. Based on data and expert knowledge, the prediction results are probability values. Therefore, supervised learning regression models are used to construct the prospecting prediction model. The machine learning models employed are artificial neural networks (MLP) and support vector machines (SVML) linear regression. Model construction consists of two steps: training and prediction. The training scope is selected by experts based on factors such as deposit type, mineralization background of the study area, research foundation of typical deposits, geochemical anomalies, linear structures, and alteration characteristics to ensure representativeness. The ratio of training data to prediction data is 3:7. After data allocation, the training data is input into the initial model to begin learning and parameter tuning. After training, the prediction data is used as input data and imported into prediction models with different parameters and algorithms for prediction, generating different prospecting potential probability maps.

[0034] Step (7) uses the ROC (Receiving Analytical Curve) to quantitatively evaluate the prediction model and determine the optimal mineral exploration prediction model. Specifically, it involves constructing a confusion matrix to comprehensively evaluate the effectiveness of the mineral exploration prediction model, and setting the prediction into two types: one is that there is mineral or no mineral in the mineral-rich grid cell, and the other is that there is mineral or no mineral in the non-mineral-rich grid cell.

[0035] The sensitivity TPR in step (7) is defined as the ratio of the number of non-mineralized grid cells incorrectly predicted as mineralized grid cells to the actual number of all non-mineralized grid cells.

[0036] The specificity FPR in step (7) is defined as the ratio of the number of correctly predicted mineralized grid cells to the actual number of all mineralized cells. The ROC curve for the prediction and evaluation of mineralized prospective areas is the ratio between the accuracy and error rate of mineral exploration prediction. The larger the AUC value, the better the comprehensive mineral exploration prediction model, and vice versa. This enables quantitative evaluation of mineral exploration prediction models with different parameters.

[0037] Step (8) uses the Yoden index to determine the threshold and delineate the distribution map of the mineralization prospect prediction area or target area. Specifically, the maximum Yoden index method is used to calculate the maximum Yoden index and its corresponding value as the threshold. Blocks below the threshold are eliminated, classified and statistically analyzed, and colored according to the size of the mineralization potential value. They are divided into two categories of prospective areas, A and B, with A being better than B. Based on the distribution map of the mineralization prospect prediction area, interactive target area delineation is carried out with experts, which greatly improves the efficiency of mineral exploration prediction.

[0038] As described above, the present invention can be well implemented. The above embodiments are only the best implementations of the present invention, but the implementation of the present invention is not limited to the above embodiments. Other changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principle of the present invention should be considered equivalent substitutions and are all included within the protection scope of the present invention.

Claims

1. A multi-data mineral exploration prediction system based on machine learning, characterized in that... The system comprises Module 1 (Ore Deposit Model Construction), Module 2 (Spatial Database Construction), Module 3 (Data Standardization), Module 4 (Predictive Model Training Data Allocation), Module 5 (Machine Learning-Based Mineral Exploration Prediction Model Construction), Module 6 (Algorithm Optimization and Parameter Tuning), and Module 7 (Potential Area Analysis and Target Area Delineation). It sequentially performs the following tasks: selecting favorable mineralization layers; comprehensively processing multi-source information and establishing a multi-source spatial database; standardizing and gridding the database data; conducting multi-model, multi-parameter mineralization potential analysis based on artificial neural networks and support vector machines; quantitatively evaluating the prediction model using ROC (Receiving Characteristic Curve) and determining the optimal mineral exploration prediction model; and using the Youden index to determine thresholds and delineate the distribution map of potential mineralization prediction areas or target areas. The working steps and conditions of each module are as follows: (1) Collect and sort out the geological data on the metallogenic geological conditions, strata, structure and magmatic activity in the study area, and analyze the geological characteristics of typical mineral deposits in the area; (2) The physical parameters in the study area are sorted out and statistically analyzed. The collected magnetic data are processed and anomaly information is extracted. The relationship between different anomaly units and known ore deposits is compared to determine the geophysical elements related to mineralization. (3) Analyze and summarize the regional geochemical characteristics and the geochemical characteristics of the study area, and use the iterative method and Youden index method to determine the lower limit of the anomaly and delineate the single-element anomaly map; use R-type cluster analysis to classify the elements, delineate two types of combined anomaly maps, and extract geochemical elements related to mineralization. (4) Image data and DEM data of the study area were collected, and mineralization, alteration information and linear structure were extracted using principal component analysis. (5) Extract geological elements related to mineralization, and establish a multi-dimensional mineralization information spatial database based on elements favorable to mineralization, and standardize and grid the data; (6) Using the multilayer perceptron (MLP) and support vector machine (SVM) machine learning models, multi-parameter mineralization potential analysis was conducted, and multiple mineral exploration prediction models were established. (7) Quantitatively evaluate multiple mineral exploration prediction models using ROC curve analysis and AUC index to determine the optimal mineral exploration prediction model; (8) Use the Yoden index to delineate the mineral exploration prediction area or mineral exploration target area based on the data of the optimal mineral exploration prediction model, and use GIS visualization spatial analysis technology and expert knowledge-driven qualitative evaluation of the mineral exploration prediction model indicators of typical mineral deposits in the study area.

2. The system according to claim 1, characterized in that: Module 1 is used to analyze the geological characteristics of mineralization and the location patterns of ore bodies; Module 2 is used to unify the scope, coordinates, and data format of the analysis data; Module 3 is used to solve the sample imbalance of the analysis data; Module 4 is used to distribute the input sample data into two categories: training data and test data; Module 5 is used to construct a mineral exploration prediction model; Module 6 is used to optimize the algorithm and adjust the parameters of the constructed mineral exploration prediction model to achieve the optimal solution; Module 7 is used to use the optimal mineral exploration prediction model to analyze potential areas and complete the target area delineation through interaction with experts.

3. The system according to claim 1, characterized in that: The step (1) of systematically collecting the exploration technical results data of the study area specifically includes: geological data, geophysical data, geochemical data, and remote sensing data.

4. The system according to claim 1 or 3, characterized in that: The geological data in step (1) includes regional metallogenic belts, stratigraphic units, magmatic rocks, tectonic-magmatic environments, metallogenic epochs, ore deposits, mineralization points, metallogenic geological bodies, veins, metallogenic structures and structural planes, ore body characteristics of typical ore deposits, mineral assemblages, ore structure and texture, and mineralization and alteration types.

5. The system according to claim 1 or 3, characterized in that: The geophysical data in step (2) are electrical data, magnetic data, and gravity data.

6. The system according to claim 1 or 3, characterized in that: The geochemical data in step (3) include river sediment measurement data, soil geochemical measurement data, and rock geochemical measurement data.

7. The system according to claim 1, characterized in that: The remote sensing data in step (4) are high-resolution image data, DEM data, and multispectral data.

8. The system according to claim 1, characterized in that: The step (5) of selecting favorable mineralization layers specifically involves: mineral deposits have their own mineralization regularity, mineralization characteristics, mineralization background, mineralization conditions, and mineralization indicators for mineralization prediction. All geophysical and remote sensing information is associated with mineral deposits and mineralization points. Factor layers favorable to mineralization are extracted, and all mineralization favorable factor information is processed by machine learning to achieve mineralization prediction and breakthroughs.

9. The system according to claim 1, characterized in that: Step (5) Extraction of favorable geochemical mineralization factors is based on the spatial distribution patterns of high and low background for chalcopyrite elements Cu, Au, Ag, Pb, Zn, Cd, As, Sb, Hg, tungsten-molybdenum group elements W, Mo, Sn, Bi, B, F, iron group elements Fe, Cr, Ni, Co, V, Ti, Mn, Mg, and rare, rare earth, and radioactive elements Li, Be, Nb, Zr, La, Y, Th, U. The lower limit of anomalies in the original geochemical data is calculated using the Youden index method and compared with the iterative estimation method. By analyzing the lower limit values ​​of anomalies obtained by the two algorithms, it is found that the lower limit value of anomalies obtained by the Youden index method is higher than that obtained by the iterative method, and the range of invalid anomalies is smaller. A better lower limit threshold is adopted. GIS software is used to interpolate and generate contour maps of each element, and R-type cluster analysis is performed to statistically identify element combinations related to the deposit type and generate element combination anomaly maps.

10. The system according to claim 1 or 9, characterized in that: The extraction of favorable factors for mineralization by remote sensing in step (5) mainly involves linear structural extraction and alteration information extraction. The remote sensing data used are Landsat 8 satellite OLI15m data, Landsat 8 satellite OLI30m data, and ASTGTM2_dem data. The alteration information is extracted directly by principal component analysis of the 30-meter resolution multispectral image. Based on the multispectral data, the information on iron staining anomaly and mud mineralization alteration is extracted. The OLI564 color composite image is formed by fusion of the 30-meter and 15-meter resolution image bands for remote sensing geological interpretation. Structural interpretation lines are extracted. The linear structural extraction uses three-dimensional scene construction and stereoscopic visualization extraction technology. The image fusion platform was ENVI5.2, and the NNDiffuse Pan Sharpeniing algorithm was used for image fusion to ensure that the fused image retains good color texture and spectral information. The resolution of the fused image reached 15 meters, and the spectral information included seven bands of short-wave infrared, near-infrared and visible light used for geological research.

11. The system according to claim 1, characterized in that: The step (5) involves the comprehensive processing of multi-source information to establish a multi-source spatial database. Specifically, this involves cropping, registering, correcting, and converting all relevant layers of favorable mineralization factors in the mineralization model and prospecting indicators to ensure that all layers have the same range, coordinates, and data format, and ultimately achieve automatic overlay and fitting of layer information.

12. The system according to claim 1 or 11, characterized in that: Step (5) database data standardization and gridding specifically involves: dividing multi-dimensional information according to the gridded block size to form unified grid block units. The gridded unit size depends on the scale, with the default block size being 50*50m. Deviation will affect the training speed of machine learning. The grid block units are assigned attribute values ​​based on whether the multi-dimensional mineralization information is covered. The mineralization information of the grid block units is distinguished by binary assignment. The location covered by mineralization information is assigned a value of 1, and the blank area is assigned a value of 0. The relationships between all layers extracted from geology, geophysics, geology, and remote sensing, as well as between layers and mineralization patterns and mineral exploration indicators, are comprehensively analyzed to determine the correlation between the standardized data and mineralization. Other redundant layers are removed. For linear structural layers that cannot be gridded, areal buffer and structural density map analysis techniques are used. The areal buffer analysis technique involves selecting buffers of different radii. A comparative study was conducted within the study area, with a default step size of 100m. The ratio of the number of mineral points to the area within the buffer zone was used as the analysis coefficient. By analyzing different buffer zone coefficients, the spatial range most significantly controlled by linear structures was determined. The structural density map analysis technique determined the grid cell size by analyzing the length and pattern of linear structures within the study area. The study area was divided into grid cells, with a default size of 100*100m. All linear structures and grid cells within the study area were overlaid and topologically reconstructed, so that each linear structure was clipped by the grid cell. The lengths of all clipped linear structures within each grid cell were summed, and the total length of linear structures within each grid cell was calculated. The structural density coefficient of all grid cells was calculated by dividing the total length of linear structures by the area of ​​each grid cell. The structural density coefficient attribute of each grid cell was extracted into points, and raster interpolation was used to generate a structural density contour map of the study area.

13. The system according to claim 1, characterized in that: The step (6) is characterized by the multi-model, multi-parameter mineralization potential analysis based on machine learning artificial neural networks and support vector machines, specifically the mineralization scale and the special characteristics of mineral exploration indicators. Based on data and expert knowledge, the prediction result is a probability value. Therefore, a supervised learning regression model is used to construct a mineral exploration prediction model. The machine learning models used are artificial neural network MLP and support vector machine SVML linear regression. The model construction is divided into two steps: training and prediction. The training range is selected by experts based on the type of ore deposit, as well as the mineralization background of the study area, the research foundation of typical ore deposits, geochemical anomalies, linear structures, and alteration characteristics to ensure representativeness. The ratio of training data to prediction data is 3:

7. After the data allocation is completed, the training data is input into the initial model to start learning and parameter tuning. After training is completed, the prediction data is used as input data and imported into prediction models with different parameters and different algorithms to make predictions and draw different mineral exploration potential probability maps.

14. The system according to claim 1, characterized in that: Step (7) uses the ROC subject characteristic curve to quantitatively evaluate the prediction model and determine the optimal mineral exploration prediction model. Specifically, a confusion matrix is ​​constructed to evaluate the overall effect of the mineral exploration prediction model. The prediction is set into two types: one is that there is mineral or no mineral in the mineral grid cell, and the other is that there is mineral or no mineral in the no-mineral grid cell.

15. The system according to claim 1 or 14, characterized in that: The sensitivity TPR in step (7) is defined as the ratio of the number of non-mineralized grid cells incorrectly predicted as mineralized grid cells to the actual number of all non-mineralized grid cells.

16. The system according to claim 1 or 14, characterized in that: The specificity FPR in step (7) is defined as the ratio of the number of correctly predicted mineralized grid cells to the actual number of all mineralized cells. The ROC curve for the prediction and evaluation of mineralized prospective areas is the ratio between the accuracy and error rate of mineral exploration prediction. The larger the AUC value, the better the comprehensive mineral exploration prediction model, and vice versa. This enables the quantitative evaluation of mineral exploration prediction models with different parameters.

17. The system according to claim 1, characterized in that: Step (8) uses the Yoden index to determine the threshold and delineate the distribution map of the mineralization prospect prediction area or target area. Specifically, the maximum Yoden index method is used to calculate the maximum Yoden index and its corresponding value as the threshold. Blocks below the threshold are eliminated, classified and statistically analyzed, and colored according to the size of the mineralization potential value. They are divided into two categories of prediction prospect areas, A and B. Category A is better than Category B. Based on the distribution map of the mineralization prospect prediction area, interactive target area delineation is carried out with experts.

Images