Mineral resource potential prediction system based on data mining

Abstract

The invention belongs to the technical field of data mining and analysis, and discloses a mineral resource potential prediction system based on data mining, comprising: collecting mineral data including geological data and chemical data; constructing a mineral resource network diagram based on the mineral data, and performing diagram analysis on the mineral resource network diagram to obtain a resource center node; carrying out network node division on the mineral resource network diagram based on the resource center node to obtain a same-level resource diagram, carrying out resource scoring on the same-level resource diagram to obtain mineral resource scores, and carrying out local feature aggregation on the same-level resource diagram to obtain resource diagram features; constructing a mineral resource potential evaluation model based on the resource map features and the mineral resource scores, and evaluating mineral resource potentials of different regions by using the mineral resource potential evaluation model; the accuracy and efficiency of mineral resource potential prediction are greatly improved, and the intellectualization and modernization of mineral resource exploration are promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of data mining and analysis, and more specifically, to a mineral resource potential prediction system based on data mining. Background Art

[0002] Against the backdrop of sustained global economic development, the demand for mineral resources continues to rise, especially in energy, construction, electronics and other fields, where the reliance on mineral resources is increasing; the development and utilization of mineral resources is an important driving force for economic development, while reasonable mineral resource assessment and potential prediction are the basis for sustainable development of the mining industry; however, traditional mineral resource potential prediction methods face many challenges, which are mainly reflected in the following aspects: many traditional mineral resource assessment methods rely mainly on the experience and judgment of geological experts, which is not only highly subjective, but also prone to bias in assessment results when faced with complex geological environments, making it difficult to accurately reflect the actual potential of resources; geological conditions are often highly complex and uncertain, and the distribution of ore bodies is often uncertain. The distribution pattern is affected by many factors, such as tectonic movement, erosion, sedimentary environment, etc. Traditional methods often lack sufficient flexibility and adaptability when dealing with such complexity. Despite the rapid development of information technology in recent years, in the field of mineral resource exploration, there are still many enterprises and institutions with a low degree of informatization, and they are unable to effectively use advanced technologies such as big data and artificial intelligence to predict resource potential. The existing mineral resource potential prediction models are mostly linear or simplified models, which make it difficult to fully capture the nonlinear relationships and complex interactions between data, resulting in insufficient prediction accuracy. The traditional mineral resource evaluation process is usually time-consuming. From data collection and analysis to the generation of the final evaluation report, it often takes months or even longer, which cannot meet the rapidly changing market needs.

[0003] In view of this, the present invention proposes a mineral resource potential prediction system based on data mining to solve the above problems. Summary of the invention

[0004] In order to overcome the above defects of the prior art and to achieve the above objectives, the present invention provides the following technical solutions: A mineral resource potential prediction system based on data mining, comprising: a data acquisition module: collecting mineral data, the mineral data including geological data and chemical data;

[0005] Resource graph module: constructs a mineral resource network graph based on mineral data, performs graph analysis on the mineral resource network graph, and obtains resource center nodes;

[0006] Resource aggregation and scoring module: including a node division unit and a feature extraction unit. The node division unit divides the network nodes of the mineral resource network diagram based on the resource center node to obtain the resource diagram of the same level, and performs resource scoring on the resource diagram of the same level to obtain the mineral resource score. The feature extraction unit aggregates the local features of the resource diagram of the same level to obtain the resource diagram features.

[0007] Model building module: Construct a mineral resource potential assessment model based on resource map characteristics and mineral resource distribution, and use the mineral resource potential assessment model to evaluate the mineral resource potential of different regions.

[0008] Furthermore, the geological data include: lithology data and structural characteristic data, and the chemical data include: ore element concentration, rare earth element concentration, pH and oxidation rate.

[0009] Furthermore, the method of constructing the mineral resource network diagram includes:

[0010] Each geological data is used as a resource node to construct a mineral resource network diagram. The resource nodes are vectorized using the one-hot encoding algorithm to obtain node vectors. The edges in the mineral resource network diagram are weighted based on the node vectors. The calculation formula for weight assignment is: Among them, Wei ab represents the resource association degree of the edge connecting the a-th resource node and the b-th resource node, Sim(a,b) represents the similarity between the node vector of the a-th resource node and the node vector of the b-th resource node, and D(a,b) represents the distance between the a-th resource node and the b-th resource node.

[0011] Furthermore, the method of performing network analysis on the mineral resource network diagram includes:

[0012] A correlation threshold is preset, and based on the a-th resource node in the mineral resource network diagram, resource nodes whose resource correlation with the edge connected to the a-th resource node is greater than or equal to the correlation threshold are taken as the correlation set of the a-th resource node; a permutation and combination algorithm is used to permutate and combine every two resource nodes in the mineral resource network diagram to obtain a resource node group, and a minimum correlation exploration is performed with one resource node in each resource node group as the starting exploration point and the other resource node as the ending exploration point to obtain a minimum correlation path; the centrality of each resource node is calculated based on the correlation set and the minimum correlation path to obtain the resource centrality of each resource node; all resource nodes are sorted from large to small according to the resource centrality of each resource node, and the first A resource nodes are selected as resource center nodes.

[0013] Furthermore, the method of performing the minimum association exploration includes:

[0014] A set of exploration points is preset, the starting exploration point is counted into the exploration point set, and the starting exploration point is marked as selected. The selection marks include selected and unselected. Selected means that the resource node has been selected, and unselected means that the resource node has not been selected. Starting from the starting exploration point, resource selection is performed on the resource nodes adjacent to the starting exploration point. Each time, the resource node with the smallest resource association with the starting exploration point and marked as unselected is selected as the stay point, and the stay point is used as the next moving position. When the stay point is reached, the resource node corresponding to the stay point is counted into the exploration point set, and the stay point is marked as selected. When the selection marks of all resource nodes adjacent to the stay point are selected, the stay point is removed from the exploration point set and returns to the previous stay point to continue resource selection. This is repeated until the end exploration point is marked as selected, and the output exploration point set at this time is the minimum association path.

[0015] Furthermore, the formula for calculating the centrality is: Among them, Cen a represents the resource centrality of the a-th resource node, L represents the number of edges directly connected to the a-th resource node in the mineral resource network graph, and Dic a Represents the associated set of the a-th resource node, Wei ac represents the resource association degree between the a-th resource node and the c-th resource node in the association set, where c represents the index of the resource node in the association set. a represents the minimum number of associated paths passing through the ath resource node, and R represents the total number of resource node groups.

[0016] Furthermore, the method of dividing the network nodes of the mineral resource network diagram includes:

[0017] Take each resource center node as the initial partition node, initialize M ants, randomly assign other resource nodes except the initial partition node to each ant, assign an initial pheromone concentration equal to the resource centrality value to each resource node based on the resource centrality of each resource node, and each ant starts from the assigned resource node and gradually selects the next resource node based on the selection probability. When the ant moves to the initial partition node, the ant stops moving. When all ants stop moving, the resource node information of the ants staying in each initial partition node is recorded as the subordinate node, and the subordinate node is aggregated with the corresponding initial partition center to obtain an aggregate network graph, and resource scoring is performed on each aggregate network graph to obtain the mineral resource score of each aggregate network graph; repeat the iteration until the mineral resource score of the aggregate network graph corresponding to each initial partition center no longer changes, and the output aggregate network graph at this time is the resource graph of the same level.

[0018] Furthermore, the formula for resource scoring is:

[0019] Among them, Gra z represents the mineral resource distribution of the zth aggregate network graph, Wall represents the sum of the resource associations corresponding to all edges in the zth aggregate network graph, Web z Represents the local node set of resource nodes paired in the zth aggregation network graph, Cen c Represents the resource centrality of the cth resource node, Cen d represents the resource centrality of the dth resource node, f(c,d) represents the score adjustment function, Sid c Represents the number of edges directly connected to the cth resource node, Sid d Represents the number of edges directly connected to the dth resource node.

[0020] Furthermore, the method of performing local feature aggregation on resource graphs of the same level includes:

[0021] Construct a spatial convolutional network, and perform local feature aggregation on resource graphs of the same level based on the spatial convolutional network to obtain resource graph features. The formula for local feature aggregation is: in, Represents the graph embedding feature of the vth peer resource graph after convolution by the Lth convolutional layer, Ad v represents the number of resource nodes in the vth peer resource graph, D represents the degree matrix of the vth peer resource graph, which is used to eliminate the impact of the imbalance of the number of connections between different nodes, and W L represents the weight matrix of the Lth convolutional layer, B L represents the bias term of the Lth convolutional layer, and σ represents the nonlinear activation function.

[0022] Furthermore, the method of constructing the mineral resource potential assessment model includes:

[0023] Based on the mineral resource score of the aggregated network graph corresponding to the resource graph features and the graph features, the graph features are used as the input of the mineral resource potential assessment model, the predicted mineral resource score is used as the output of the mineral resource potential assessment model, and the actual mineral resource score is used as the prediction target of the mineral resource potential assessment model for model training to construct a mineral resource potential assessment model. The mineral resource potential assessment model is a linear regression model, and the target loss function of the mineral resource potential assessment model is a mean square error function. Minimizing the value of the target loss function is used as the training goal to obtain a mineral resource potential assessment model with the smallest value of the target loss function.

[0024] The technical effects and advantages of the mineral resource potential prediction system based on data mining of the present invention are as follows:

[0025] The present invention vectorizes lithology, structural features, etc. through one-hot encoding, so that the resource nodes in the mineral resource network diagram can effectively describe the geological and chemical characteristics of different mining areas, which is helpful to deeply explore the potential laws in the data; by performing node correlation analysis and minimum correlation path exploration on the mineral resource network diagram, it is helpful to identify potential concentrated areas of mineral resources, discover areas with high correlation and potential in geological structures, and improve the accuracy of potential prediction; through the division method based on the ant algorithm, not only the accuracy of resource concentration areas is improved, but also complex geological conditions can be adapted; through the spatial convolutional network, convolution operations are performed on resource maps of the same level, and the system can capture the spatial dependence of resource nodes, and then extract the spatial characteristics of mineral resource distribution, further improving the accuracy of prediction; by combining the mineral resource division and the features extracted by the spatial convolutional network, complex nonlinear relationships can be captured more accurately, and the evaluation accuracy can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a schematic diagram of a mineral resource potential prediction system based on data mining of the present invention;

[0027] Figure 2 It is a schematic diagram of the mineral resource potential prediction method based on data mining of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] Example 1

[0030] See also Figure 1 As shown, the mineral resource potential prediction system based on data mining described in this embodiment includes: a data acquisition module: collecting mineral data, the mineral data includes geological data and chemical data;

[0031] Resource graph module: constructs a mineral resource network graph based on mineral data, performs graph analysis on the mineral resource network graph, and obtains resource center nodes;

[0032] Resource aggregation and scoring module: including a node division unit and a feature extraction unit. The node division unit divides the network nodes of the mineral resource network diagram based on the resource center node to obtain the resource diagram of the same level, and performs resource scoring on the resource diagram of the same level to obtain the mineral resource score. The feature extraction unit aggregates the local features of the resource diagram of the same level to obtain the resource diagram features.

[0033] Model building module: construct a mineral resource potential assessment model based on resource map characteristics and mineral resource distribution, and use the mineral resource potential assessment model to assess the mineral resource potential of different regions;

[0034] The modules are connected to each other via wired and / or wireless means to achieve data transmission between modules.

[0035] Geological data include: lithology data and structural characteristic data, and chemical data include: ore element concentration, rare earth element concentration, pH and oxidation rate; lithology data is used to describe the lithology of the strata, such as granite, basalt and shale, etc., and lithology data directly affects the type and distribution of minerals; structural characteristic data is used to describe the geological structure, such as faults and joints, etc. Structural characteristics are often an important factor in the formation of mineral deposits, especially in sedimentary and volcanic deposits; ore element concentration is the concentration of metal elements (such as copper, iron, and nickel), and the higher the concentration area, the higher the potential; rare earth element concentration is the concentration of rare earth elements (such as lanthanum and cerium), and rare earth elements have special indicative significance for certain specific types of deposits; pH and oxidation rate are the chemical environment of the mining area, which are used to assist in identifying mineral components.

[0036] Each geological data is used as a resource node to construct a mineral resource network diagram. The resource nodes are vectorized using the one-hot encoding algorithm to obtain node vectors. The edges in the mineral resource network diagram are weighted based on the node vectors. The calculation formula for weight assignment is: Among them, Wei ab represents the resource association degree of the edge connecting the a-th resource node and the b-th resource node, Sim(a,b) represents the similarity between the node vector of the a-th resource node and the node vector of the b-th resource node, and the commonly used similarity measurement formulas include cosine similarity calculation, Manhattan distance calculation, and Chebyshev distance calculation. D(a,b) represents the distance between the a-th resource node and the b-th resource node, and the commonly used distance calculation formulas include the Euclidean distance formula; assuming that the lithology data includes granite, basalt, and shale, the way to perform unique hot encoding on the lithology data is: define the reference value of granite as 0, the reference value of basalt as 1, and the reference value of shale as 2. When the lithology data in a certain geological data is granite, the value of the lithology data feature in its corresponding node vector is 0;

[0037] The methods of conducting network analysis on mineral resource network diagrams include:

[0038] A correlation threshold is preset, and based on the a-th resource node in the mineral resource network diagram, resource nodes whose resource correlation with the edges connected to the a-th resource node is greater than or equal to the correlation threshold are taken as the correlation set of the a-th resource node; a permutation and combination algorithm is used to permutate and combine every two resource nodes in the mineral resource network diagram to obtain a resource node group, and a resource node in each resource node group is used as a starting exploration point, and the other resource node is used as an ending exploration point to perform minimum correlation exploration. The method of performing minimum correlation exploration includes: presetting an exploration point set, counting the starting exploration point into the exploration point set, and selecting and marking the starting exploration point as "selected", and the selection marks include "selected" and "unselected", "selected" means that the resource node has been selected, and "unselected" means that the resource node has not been selected, starting from the starting exploration point, the resources adjacent to the starting exploration point are selected. The resource node selects resources. Each time, the resource node with the smallest resource association with the starting exploration point and marked as "unselected" is selected as the stop point, and the stop point is used as the next moving position. When the stop point is reached, the resource node corresponding to the stop point is counted into the exploration point set, and the stop point is marked as "selected". When the selection marks of all resource nodes adjacent to the stop point are "selected", the stop point is removed from the exploration point set and returns to the previous stop point to continue resource selection. This is repeated until the end exploration point is marked as "selected". The exploration point set output at this time is the minimum association path. The selection of the path based on the greedy algorithm ensures that the selected nodes will not be visited repeatedly during the exploration process, reducing unnecessary calculations. Dynamically updating the mark not only ensures the uniqueness of the path, but also reduces the complexity of the algorithm, making the path exploration more efficient.

[0039] The centrality of each resource node is calculated based on the association set and the minimum association path. The formula for centrality calculation is: Among them, Cen a represents the resource centrality of the a-th resource node, L represents the number of edges directly connected to the a-th resource node in the mineral resource network graph, and Dic a Represents the associated set of the a-th resource node, Wei ac represents the resource association degree between the a-th resource node and the c-th resource node in the association set, where c represents the index of the resource node in the association set. a represents the minimum number of associated paths passing through the ath resource node, and R represents the total number of resource node groups; all resource nodes are sorted from large to small according to the resource centrality of each resource node, and the first A resource nodes are selected as resource central nodes.

[0040] The network nodes of the mineral resource network diagram are divided based on the resource center node, and the resource center node is selected as the initial division node of each area. This can ensure that the network is divided from the position with the strongest resource correlation in the network, improve the rationality of the division, and avoid irrelevant areas from being classified into the clustered area. The method of dividing the network nodes includes: taking each resource center node as the initial division node, initializing M ants, M is an integer greater than zero, and randomly assigning other resource nodes except the initial division node to each ant. Based on the resource centrality of each resource node, an initial pheromone concentration of the size of the resource centrality value is assigned to each resource node. Each ant starts from the assigned resource node and gradually selects the next resource node based on the selection probability. The calculation formula of the selection probability is: Among them, P ij represents the probability of an ant moving from the i-th resource node to the j-th resource node, ΔND ij Represents the difference in initial pheromone concentration between the i-th resource node and the j-th resource node, Wei ij represents the resource association between the i-th resource node and the j-th resource node, N represents the set of resource nodes that have a direct edge with the i-th resource node, and the ant stops moving when it moves to the initial partition node. When all ants stop moving, the resource node information of the ant staying in each initial partition node is recorded as the subordinate node, and the subordinate node is aggregated with the corresponding initial partition center to obtain an aggregated network graph, and resource scoring is performed on each aggregated network graph. The formula for resource scoring is: Among them, Gra z represents the mineral resource distribution of the zth aggregate network graph, Wall represents the sum of the resource associations corresponding to all edges in the zth aggregate network graph, Web z Represents the local node set of resource nodes paired in the zth aggregation network graph, Cen c Represents the resource centrality of the cth resource node, Cen d represents the resource centrality of the d-th resource node, f(c,d) represents the score adjustment function, which is used to expand the influence of directly connected resource nodes on the mineral resource score. When the c-th resource node is directly connected to the d-th resource node, the value of the score adjustment function is 1. When the c-th resource node is not directly connected to the d-th resource node, the value of the score adjustment function is 0.1. Sid c Represents the number of edges directly connected to the cth resource node, Sid dRepresents the number of edges directly connected to the dth resource node; repeat the iteration until the mineral resource score of the aggregation network graph corresponding to each initial division center no longer changes, and the output aggregation network graph at this time is the same-level resource graph; each same-level resource graph represents a potential mineral resource aggregation area, and these areas are similar in geological characteristics. By analyzing the resource nodes in these same-level resource graphs through resource scoring, the mineral resource potential of each area can be evaluated. The higher the mineral resource score, the greater the mineral resource potential, and the more likely it is to contain rich mineral resources.

[0041] Construct a spatial convolutional network, and perform local feature aggregation on resource graphs of the same level based on the spatial convolutional network to obtain resource graph features. The spatial convolutional layer of the spatial convolutional network is used to perform convolution operations on resource graphs of different levels to capture the spatial dependencies of different resource nodes. The formula for local feature aggregation is:

[0042] in, Represents the graph embedding feature of the vth peer resource graph after convolution by the Lth convolutional layer, Ad v represents the number of resource nodes in the vth peer resource graph, D represents the degree matrix of the vth peer resource graph, which is used to eliminate the impact of the imbalance of the number of connections between different nodes, and W L represents the weight matrix of the Lth convolutional layer, B L represents the bias term of the Lth convolutional layer, σ represents the nonlinear activation function. Commonly used nonlinear activation functions include ReLU function, Sigmoid function and Swish function. Through network node division, resource scoring and local feature aggregation, the system can efficiently identify the potential scores of different mining areas. The scoring adjustment function combines the influence of directly connected resource nodes to reflect the distribution characteristics of mineral resources within and between regions, ensuring the accuracy of mineral resource potential assessment; the degree matrix is ​​a diagonal matrix, and its diagonal elements represent the degrees of the corresponding resource nodes, that is, the number of edges connected to the resource nodes. Assume that the degree of node 1 is 2 (connecting node 2 and node 3), the degree of node 2 is 2 (connecting node 1 and node 3), and the degree of node 3 is 1 (connecting node 1). Therefore, the degree matrix D is:

[0043] Based on the mineral resource score of the resource map features and the aggregated network graph corresponding to the resource map features, the graph features are used as the input of the mineral resource potential assessment model, the predicted mineral resource score is used as the output of the mineral resource potential assessment model, and the actual mineral resource score is used as the prediction target of the mineral resource potential assessment model for model training. The mineral resource potential assessment model is constructed. The mineral resource potential assessment model is a logistic regression model, and the target loss function of the mineral resource potential assessment model is: Among them, MSE represents the mean square error of potential, Su represents the total amount of data of resource map features, and F g represents the g-th mineral resource, F' g Represents the g-th predicted mineral resource score; taking the minimization of the value of the target loss function as the training goal, a mineral resource potential assessment model with the smallest value of the target loss function is obtained, and the mineral resource potential of different regions is predicted based on the trained mineral resource potential assessment model to visualize the mineral resource potential of different regions.

[0044] This embodiment uses unique-hot encoding to vectorize lithology, structural characteristics, etc., so that the resource nodes in the mineral resource network diagram can effectively describe the geological and chemical characteristics of different mining areas, which is helpful to deeply explore the potential laws in the data; by performing correlation analysis and minimum correlation path exploration on the nodes of the mineral resource network diagram, it is helpful to identify potential concentration areas of mineral resources, discover areas with high correlation and potential in the geological structure, and improve the accuracy of potential prediction; through the division method based on the ant algorithm, not only the accuracy of resource concentration areas is improved, but also it can adapt to complex geological conditions; through the spatial convolutional network to perform convolution operations on the resource maps of the same level, the system can capture the spatial dependence of resource nodes, and then extract the spatial characteristics of mineral resource distribution, and further improve the accuracy of prediction; by combining the mineral resource division and the features extracted by the spatial convolutional network, it can more accurately capture complex nonlinear relationships and improve the evaluation accuracy.

[0045] Example 2

[0046] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, and a method for predicting mineral resource potential based on data mining is provided, including:

[0047] S1. Collect mineral data, which includes geological data and chemical data;

[0048] S2. Construct a mineral resource network diagram based on mineral data, perform graph analysis on the mineral resource network diagram, and obtain resource center nodes;

[0049] S3. Divide the network nodes of the mineral resource network diagram based on the resource center node to obtain the resource diagram of the same level, perform resource scoring on the resource diagram of the same level to obtain the mineral resource score, perform local feature aggregation on the resource diagram of the same level to obtain the resource diagram feature;

[0050] S4. Construct a mineral resource potential assessment model based on resource map characteristics and mineral resource distribution, and use the mineral resource potential assessment model to assess the mineral resource potential of different regions.

[0051] Example 3

[0052] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the mineral resource potential prediction method based on data mining provided above is implemented.

[0053] Since the electronic device introduced in this embodiment is an electronic device used to implement the method for predicting the potential of mineral resources based on data mining in the embodiment of the present application, based on the method for predicting the potential of mineral resources based on data mining introduced in the embodiment of the present application, those skilled in the art can understand the specific implementation of the electronic device of the present embodiment and its various variations, so how the electronic device implements the method in the embodiment of the present application is not described in detail here. As long as those skilled in the art implement the electronic device used by the method for predicting the potential of mineral resources based on data mining in the embodiment of the present application, they all belong to the scope to be protected by this application.

[0054] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0055] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. The mineral resource potential prediction system based on data mining is characterized by: include: Data acquisition module: collects mineral data, including geological data and chemical data; Resource graph module: construct a mineral resource network graph based on mineral data, perform graph analysis on the mineral resource network graph, and obtain resource center nodes; Resource aggregation and scoring module: including a node division unit and a feature extraction unit. The node division unit divides the network nodes of the mineral resource network diagram based on the resource center node to obtain the resource diagram of the same level, and performs resource scoring on the resource diagram of the same level to obtain the mineral resource score. The feature extraction unit aggregates the local features of the resource diagram of the same level to obtain the resource diagram features. Model building module: Construct a mineral resource potential assessment model based on resource map characteristics and mineral resource distribution, and use the mineral resource potential assessment model to evaluate the mineral resource potential of different regions.

2. The mineral resource potential prediction system based on data mining according to claim 1 is characterized in that: The geological data include: lithology data and structural characteristic data, and the chemical data include: ore element concentration, rare earth element concentration, pH and oxidation rate.

3. The mineral resource potential prediction system based on data mining according to claim 2 is characterized in that: The method of constructing the mineral resource network diagram includes: Each geological data is used as a resource node to construct a mineral resource network diagram. The resource nodes are vectorized using the one-hot encoding algorithm to obtain node vectors. The edges in the mineral resource network diagram are weighted based on the node vectors. The calculation formula for weight assignment is: Among them, Wei ab represents the resource association degree of the edge connecting the a-th resource node and the b-th resource node, Sim(a,b) represents the similarity between the node vector of the a-th resource node and the node vector of the b-th resource node, and D(a,b) represents the distance between the a-th resource node and the b-th resource node.

4. The mineral resource potential prediction system based on data mining according to claim 3 is characterized in that: The method of performing network analysis on the mineral resource network diagram includes: A correlation threshold is preset, and based on the a-th resource node in the mineral resource network diagram, resource nodes whose resource correlation with the edge connected to the a-th resource node is greater than or equal to the correlation threshold are taken as the correlation set of the a-th resource node; a permutation and combination algorithm is used to permutate and combine every two resource nodes in the mineral resource network diagram to obtain a resource node group, and a minimum correlation exploration is performed with one resource node in each resource node group as the starting exploration point and the other resource node as the ending exploration point to obtain a minimum correlation path; the centrality of each resource node is calculated based on the correlation set and the minimum correlation path to obtain the resource centrality of each resource node; all resource nodes are sorted from large to small according to the resource centrality of each resource node, and the first A resource nodes are selected as resource center nodes.

5. The mineral resource potential prediction system based on data mining according to claim 4 is characterized in that: The method of performing the minimum association exploration includes: A set of exploration points is preset, the starting exploration point is counted into the exploration point set, and the starting exploration point is marked as selected. The selection marks include selected and unselected. Selected means that the resource node has been selected, and unselected means that the resource node has not been selected. Starting from the starting exploration point, resource selection is performed on the resource nodes adjacent to the starting exploration point. Each time, the resource node with the smallest resource association with the starting exploration point and marked as unselected is selected as the stay point, and the stay point is used as the next moving position. When the stay point is reached, the resource node corresponding to the stay point is counted into the exploration point set, and the stay point is marked as selected. When the selection marks of all resource nodes adjacent to the stay point are selected, the stay point is removed from the exploration point set and returns to the previous stay point to continue resource selection. This is repeated until the end exploration point is marked as selected, and the output exploration point set at this time is the minimum association path.

6. The mineral resource potential prediction system based on data mining according to claim 5 is characterized in that: The formula for calculating centrality is: Among them, Cen a represents the resource centrality of the a-th resource node, L represents the number of edges directly connected to the a-th resource node in the mineral resource network graph, and Dic a Represents the associated set of the a-th resource node, Wei ac represents the resource association degree between the a-th resource node and the c-th resource node in the association set, where c represents the index of the resource node in the association set. a represents the minimum number of associated paths passing through the ath resource node, and R represents the total number of resource node groups.

7. The mineral resource potential prediction system based on data mining according to claim 6 is characterized in that: The method of dividing the network nodes of the mineral resource network diagram includes: Take each resource center node as the initial partition node, initialize M ants, randomly assign other resource nodes except the initial partition node to each ant, assign an initial pheromone concentration equal to the resource centrality value to each resource node based on the resource centrality of each resource node, and each ant starts from the assigned resource node and gradually selects the next resource node based on the selection probability. When the ant moves to the initial partition node, the ant stops moving. When all ants stop moving, the resource node information of the ants staying in each initial partition node is recorded as the subordinate node, and the subordinate node is aggregated with the corresponding initial partition center to obtain an aggregate network graph, and resource scoring is performed on each aggregate network graph to obtain the mineral resource score of each aggregate network graph; repeat the iteration until the mineral resource score of the aggregate network graph corresponding to each initial partition center no longer changes, and the output aggregate network graph at this time is the resource graph of the same level.

8. The mineral resource potential prediction system based on data mining according to claim 7 is characterized in that: The formula for resource scoring is: Among them, Gra z represents the mineral resource distribution of the zth aggregate network graph, Wall represents the sum of the resource associations corresponding to all edges in the zth aggregate network graph, Web z Represents the local node set of resource nodes paired in the zth aggregation network graph, Cen c Represents the resource centrality of the cth resource node, Cen d represents the resource centrality of the dth resource node, f(c,d) represents the score adjustment function, Sid c Represents the number of edges directly connected to the cth resource node, Sid d Represents the number of edges directly connected to the dth resource node.

9. The mineral resource potential prediction system based on data mining according to claim 8 is characterized in that: The method of performing local feature aggregation on resource graphs of the same level includes: Construct a spatial convolutional network, and perform local feature aggregation on resource graphs of the same level based on the spatial convolutional network to obtain resource graph features. The formula for local feature aggregation is: in, Represents the graph embedding feature of the vth peer resource graph after convolution by the Lth convolutional layer, Ad v represents the number of resource nodes in the vth peer resource graph, D represents the degree matrix of the vth peer resource graph, which is used to eliminate the impact of the imbalance of the number of connections between different nodes, and W L represents the weight matrix of the Lth convolutional layer, B L represents the bias term of the Lth convolutional layer, and σ represents the nonlinear activation function.

10. The mineral resource potential prediction system based on data mining according to claim 9, characterized in that: The method of constructing the mineral resource potential assessment model includes: Based on the mineral resource score of the aggregated network graph corresponding to the resource graph features and the graph features, the graph features are used as the input of the mineral resource potential assessment model, the predicted mineral resource score is used as the output of the mineral resource potential assessment model, and the actual mineral resource score is used as the prediction target of the mineral resource potential assessment model for model training to construct a mineral resource potential assessment model. The mineral resource potential assessment model is a linear regression model, and the target loss function of the mineral resource potential assessment model is a mean square error function. Minimizing the value of the target loss function is used as the training goal to obtain a mineral resource potential assessment model with the smallest value of the target loss function.

Images