Deposit reserve prediction method, training method of prediction model

Abstract

The application provides a deposit reserve prediction method and a training method of a prediction model, and can be applied to the technical field of deep learning. The deposit reserve prediction method comprises the following steps: performing feature coding on deposit information of a to-be-predicted deposit and ore information used for the to-be-predicted deposit respectively to obtain deposit features and ore features; performing feature segmentation on the deposit features and the ore features according to a preset feature segmentation dimension to obtain a plurality of deposit sub-features and a plurality of ore sub-features; randomly determining one sub-feature from the plurality of deposit sub-features and the plurality of ore sub-features respectively for splicing to obtain first splicing features; and inputting the first splicing features into a prediction model to output target deposit reserves of the to-be-predicted deposit.

Description

Technical Field

[0001] This invention relates to the field of deep learning technology, and more specifically to a method for predicting mineral deposit reserves and a method for training a prediction model. Background Technology

[0002] As mining operations commence, accurately assessing ore reserves is crucial for mining companies to make informed investment decisions and achieve higher economic benefits and returns on investment. Related technologies involve constructing three-dimensional ore deposit models based on geological data obtained from on-site surveys, and then using these models to infer the amount of ore reserves.

[0003] In realizing the concept of this invention, the related technologies have at least the following problems: the spatial distribution of mineral deposits in nature is very complex. At present, we are only looking for suitable simple polyhedra to replace the mineral-bearing geological bodies in the real world. This process requires a lot of geological knowledge. Then, the distribution of ore bodies in two-dimensional space is calculated by humans and then extrapolated to three-dimensional space for calculation. The calculated mineral reserves are easily affected by human factors, making the volume formula of the mineral-bearing geological body inaccurate and reducing the accuracy of mineral reserve estimation. Summary of the Invention

[0004] In view of the above problems, the present invention provides a mineral deposit reserve prediction method and a prediction model training method to improve the accuracy of mineral deposit reserve prediction.

[0005] According to a first aspect of the present invention, a method for predicting mineral deposit reserves is provided. The method includes: performing feature encoding on mineral deposit information and ore information for the mineral deposit to be predicted, respectively, to obtain mineral deposit features and ore features; performing feature segmentation on the mineral deposit features and ore features according to a preset feature segmentation dimension, respectively, to obtain multiple mineral deposit sub-features and multiple ore sub-features; randomly selecting one sub-feature from each of the multiple mineral deposit sub-features and multiple ore sub-features for splicing, to obtain a first spliced ​​feature; inputting the first spliced ​​feature into a prediction model, and outputting the target mineral deposit reserves of the mineral deposit to be predicted.

[0006] According to an embodiment of the present invention, the prediction model includes a self-attention module and an evaluation module; the step of inputting the first splicing feature into the prediction model and outputting the target mineral deposit reserve of the mineral deposit to be predicted includes: inputting the first splicing feature into the self-attention module and outputting an interaction feature; inputting the interaction feature into the evaluation module so that the evaluation module evaluates multiple candidate mineral deposit reserves based on the interaction feature and outputs the target mineral deposit reserve, wherein the multiple candidate mineral deposit reserves are determined during the training of the prediction model.

[0007] According to an embodiment of the present invention, the self-attention module includes a plurality of attention sub-modules; the step of inputting the first splicing feature into the self-attention module and outputting the interaction feature includes: inputting the first splicing feature into the plurality of attention sub-modules respectively and outputting a plurality of interaction sub-features; splicing the plurality of interaction sub-features to obtain a second splicing feature; and reconstructing the second splicing feature to obtain the interaction feature.

[0008] According to an embodiment of the present invention, the above-mentioned feature reconstruction of the second splicing feature to obtain the interactive feature includes: performing a linear transformation on the second splicing feature according to a linear transformation function to obtain a transformed feature; and performing a nonlinear reconstruction on the transformed feature according to a nonlinear transformation function to obtain the interactive feature.

[0009] According to an embodiment of the present invention, the above-mentioned input of the interaction features into the evaluation module so that the evaluation module evaluates the reserves of multiple candidate mineral deposits based on the interaction features and outputs the target mineral deposit reserves includes: when the evaluation module receives the interaction features, performing a nonlinear transformation on the interaction features to enhance the semantic expressive power of the interaction features and obtain enhanced feature values; determining the target mineral deposit reserves among the multiple candidate mineral deposit reserves according to the matching degree between the enhanced feature values ​​and the candidate reserve quantum features of each of the multiple candidate mineral deposit reserves, and outputting the target mineral deposit reserves.

[0010] According to an embodiment of the present invention, determining the target mineral deposit reserve among the multiple candidate mineral deposit reserves based on the matching degree between the enhanced feature value and the candidate reservoir quantum features of each of the multiple candidate mineral deposit reserves includes: for each candidate mineral deposit reserve, determining a target attribute feature value with the same sampling depth as the ore information from multiple initial attribute feature values, wherein the initial attribute feature value is the product of the mineral deposit sample features and the ore sample features used in the training process; and determining the target attribute feature value based on the mineral deposit sub-features, the ore sub-features, the candidate reservoir quantum features, and the ore density gradient. The potential feature value is determined by multiplying the values ​​of the potential feature value, the target attribute feature value, and the reinforcement feature value obtained in each training round. Based on the multiple weight parameters obtained in each training round, the potential feature value, the target attribute feature value, and the reinforcement feature value of each training round are weighted and summed to obtain a first evaluation value. The second evaluation value is determined by multiplying the deposit sub-feature, the ore sub-feature, and the candidate reservoir quantum feature, and the first evaluation value is determined to match the reinforcement feature value with the candidate reservoir quantum feature of the candidate deposit reserve. Based on the matching degree of each candidate deposit reserve, the target deposit reserve is determined among the multiple candidate deposit reserves.

[0011] According to an embodiment of the present invention, the above-mentioned deposit information includes the deposit name and deposit thickness, and the above-mentioned ore information includes the sampling depth and the ore density corresponding to the above-mentioned sampling depth.

[0012] According to an embodiment of the present invention, the above-mentioned deposit features include deposit name features and deposit thickness features, and the above-mentioned ore features include sampling depth features and ore density features; the above-mentioned feature segmentation of the above-mentioned deposit features and the above-mentioned ore features according to a preset feature segmentation dimension to obtain multiple deposit sub-features and multiple ore sub-features includes: horizontally splicing the above-mentioned sampling depth features and the above-mentioned ore density features to obtain a third spliced ​​feature; and feature segmentation of the above-mentioned deposit name features, the above-mentioned deposit thickness features and the above-mentioned third spliced ​​feature according to a preset feature segmentation dimension to obtain multiple name sub-features, multiple thickness sub-features and the above-mentioned multiple ore sub-features.

[0013] According to an embodiment of the present invention, the above method further includes: extracting data from the above-mentioned mineral deposit to be predicted based on the mineral deposit dataset to obtain an initial mineral deposit name; and standardizing the above-mentioned initial mineral deposit name according to a preset naming template to obtain the above-mentioned mineral deposit name.

[0014] A second aspect of the present invention provides a method for training a prediction model. The method includes: randomly determining multiple sample sets from sample sets of multiple mineral deposits, wherein the sample sets include multiple sample data of the same mineral deposit, and the sample data includes sample mineral deposit information, sample ore information, and sample reserve information; performing feature encoding on the sample mineral deposit information, the sample ore information, and the sample reserve information respectively to obtain sample mineral deposit features, sample ore features, and sample reserve features; performing feature segmentation on the sample mineral deposit features and the sample ore features according to a preset feature segmentation dimension to obtain multiple sample mineral deposit sub-features and multiple sample ore sub-features; randomly determining one sub-feature from each of the multiple sample mineral deposit sub-features and the multiple sample ore sub-features for feature concatenation to obtain a first sample concatenation feature; inputting the first sample concatenation feature into an initial prediction model to output a predicted reserve feature; calculating the loss value of the initial prediction model based on the sample reserve feature and the predicted reserve feature; repeating the above operations until the loss value meets a preset condition to obtain a prediction model.

[0015] According to an embodiment of the present invention, the deposit features and ore features are first segmented to obtain multiple deposit sub-features and multiple ore sub-features. Then, a sub-feature is randomly selected from each of these deposit sub-features and ore sub-features and spliced ​​together to obtain a first spliced ​​feature. The target deposit reserve is determined using the first spliced ​​feature, which reduces the probability of overfitting of the prediction model due to excessive feature data input to the prediction model, thereby improving the accuracy of the target deposit reserve. The first spliced ​​feature obtained by splicing the randomly selected deposit sub-features and ore sub-features can construct explicit feature interaction terms, which helps the prediction model capture features that are highly correlated with the predicted target deposit reserve in the first spliced ​​feature, further improving the accuracy of the target deposit reserve. Attached Figure Description

[0016] The above-mentioned contents, as well as other objects, features and advantages of the present invention, will become clearer from the following description of embodiments of the present invention with reference to the accompanying drawings.

[0017] Figure 1 The diagram illustrates an application scenario of the mineral deposit reserve prediction method and the prediction model training method according to an embodiment of the present invention.

[0018] Figure 2 A flowchart of a mineral deposit reserve prediction method according to an embodiment of the present invention is shown.

[0019] Figure 3 A schematic diagram of feature segmentation according to an embodiment of the present invention is shown.

[0020] Figure 4 A schematic diagram of a mineral deposit reserve prediction method according to an embodiment of the present invention is shown.

[0021] Figure 5 A flowchart illustrating a training method for a prediction model according to an embodiment of the present invention is shown.

[0022] Figure 6 A structural block diagram of a mineral deposit reserve prediction device according to an embodiment of the present invention is shown.

[0023] Figure 7 A structural block diagram of a training apparatus for a prediction model according to an embodiment of the present invention is shown.

[0024] Figure 8 A block diagram of an electronic device suitable for implementing a mineral deposit reserve prediction method according to an embodiment of the present invention is shown. Detailed Implementation

[0025] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the invention. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the invention for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0026] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0027] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0028] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0029] In the technical solution of this invention, the collection, updating, analysis, processing, use, transmission, provision, disclosure, and storage of data (e.g., including but not limited to user personal information) comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures have been taken to prevent unauthorized access to user personal information data and to maintain the security of user personal information and network security.

[0030] Mineral deposits are a significant component of mining costs. Obtaining accurate information on mineral reserves can help companies rationally plan their mining operations to achieve higher economic benefits and realize the sustainable exploitation and utilization of resources.

[0031] In related technologies, resource estimation methods are used to predict the potential of mineral resources. These methods mainly include the three-step method, volume method, abundance value method, deposit model method, subjective probability method, and grade-tonnage model method. However, these methods generally suffer from the following problems: the spatial distribution of ore bodies in nature is extremely complex. Currently, we are only looking for suitable simple polyhedra to represent real-world ore-bearing geological bodies. This process requires extensive geological knowledge to determine the ore distribution in two-dimensional space before extrapolating to three-dimensional space for calculation. The determination of the ore distribution in two-dimensional space is easily affected by differences in individual geological knowledge, leading to inaccurate volume formulas for ore-bearing geological bodies and reducing the accuracy of the final determined ore deposit reserves.

[0032] An embodiment of the present invention provides a method for predicting mineral deposit reserves, comprising: performing feature encoding on mineral deposit information and ore information for the mineral deposit to be predicted, respectively, to obtain mineral deposit features and ore features; performing feature segmentation on the mineral deposit features and ore features according to a preset feature segmentation dimension, respectively, to obtain multiple mineral deposit sub-features and multiple ore sub-features; randomly selecting one sub-feature from each of the multiple mineral deposit sub-features and multiple ore sub-features for splicing, to obtain a first spliced ​​feature; inputting the first spliced ​​feature into a prediction model, and outputting the target mineral deposit reserves of the mineral deposit to be predicted.

[0033] Figure 1 The diagram illustrates an application scenario of the mineral deposit reserve prediction method and the prediction model training method according to an embodiment of the present invention.

[0034] like Figure 1 As shown, application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0035] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0036] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0037] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0038] It should be noted that the mineral deposit reserve prediction method and prediction model training method provided in the embodiments of the present invention can generally be executed by server 105. The mineral deposit reserve prediction method and prediction model training method provided in the embodiments of the present invention can also be executed by a server or server cluster that is different from server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or server 105.

[0039] It should be understood that Figure 1 The number of first terminal devices, second terminal devices, third terminal devices, network terminal devices, networks, and servers shown in the diagram is merely illustrative. Depending on implementation needs, any number of first terminal devices, second terminal devices, third terminal devices, networks, and servers can be included.

[0040] The following will be based on Figure 1 The described scene, through Figures 2-5 The mineral deposit reserve prediction method and prediction model training method of the present invention are described in detail.

[0041] Figure 2 A flowchart of a mineral deposit reserve prediction method according to an embodiment of the present invention is shown.

[0042] like Figure 2 As shown, the mineral deposit reserve prediction method in this embodiment includes operations S210 to S240.

[0043] In operation S210, the deposit information of the deposit to be predicted and the ore information used for the deposit to be predicted are respectively feature-encoded to obtain deposit features and ore features.

[0044] In operation S220, the deposit features and ore features are segmented according to the preset feature segmentation dimensions to obtain multiple deposit sub-features and multiple ore sub-features.

[0045] In operation S230, one sub-feature is randomly selected from multiple ore deposit sub-features and multiple ore sub-features for splicing to obtain the first spliced ​​feature.

[0046] In operation S240, the first splicing feature is input into the prediction model, and the target mineral deposit reserves of the deposit to be predicted are output.

[0047] According to embodiments of the present invention, the deposit information is information representing the deposit to be predicted in the deposit dimension. For example, the deposit information may include the geographical location of the deposit, deposit number, etc.; the ore information is information representing the deposit to be predicted in the ore dimension. For example, the ore information may include ore density, ore name, etc. The ore corresponding to the ore information belongs to the deposit to be predicted. For example, the ore information may be the ore density of the deposit to be predicted at a preset depth.

[0048] The deposit information and ore information used to describe the deposit to be predicted are respectively feature-encoded to obtain deposit features and ore features, so that the deposit information and ore information can be embedded into the feature matrix or feature vector that can be processed by the prediction model.

[0049] The preset feature segmentation dimension can be a dimension that allows the deposit sub-features and ore sub-features to contain as much information as possible across multiple dimensions. For example, deposit information includes information 1 and information 2, where information 1 can be the geographical location of the deposit to be predicted, and information 2 can be the deposit number of the deposit to be predicted. Feature encoding of information 1 yields matrix A, and feature encoding of information 2 yields matrix B. The encoded deposit features are matrix AB. The preset feature segmentation dimension is the matrix row dimension. Feature segmentation of the deposit features along the matrix row dimension yields multiple deposit sub-features. Each deposit sub-feature includes partial information from both matrix A and matrix B. The partial information from matrix A in each deposit sub-feature collectively constitutes matrix A, and the partial information from matrix B in each deposit sub-feature collectively constitutes matrix B. This feature segmentation method is also applicable to feature segmentation of ore features.

[0050] Feature segmentation of deposit and ore features can reduce the correlation between sub-features and increase the correlation within sub-features, thus reducing the possibility of decreased accuracy of prediction model output due to the separation of strongly interactive features.

[0051] The first splicing feature includes ore deposit sub-features and ore sub-features to ensure the richness of features in the first splicing feature. It can be understood that there is no necessary feature correspondence between multiple ore deposit sub-features and multiple ore sub-features. Therefore, the first splicing feature obtained by splicing randomly determined ore deposit sub-features and ore sub-features is used to objectively reflect the information of the combined sub-features. This free combination will not introduce information that did not originally exist or misleading information.

[0052] For example, if there are multiple ore deposit sub-features including ore deposit sub-feature 1 and ore deposit sub-feature 2, and multiple ore sub-features including ore sub-feature 1 and ore sub-feature 2, then ore deposit sub-feature 1 can be combined with ore sub-feature 1 or ore sub-feature 2.

[0053] The process of splicing randomly determined deposit sub-features and ore sub-features to obtain the first spliced ​​feature can be represented by formula (1):

[0054] Formula (1);

[0055] in, Indicates the sub-characteristics of the ore deposit, Indicates the characteristics of ore particles. Indicates the first splicing feature. Represents the logarithmic function. This represents the concatenation function.

[0056] The prediction model is used to calculate the target mineral reserves of the mineral deposit to be predicted based on the information represented in the first spliced ​​feature.

[0057] Understandably, the greater the amount of information in the deposit and ore information, the larger the amount of data in the encoded deposit and ore features. Since the first concatenated feature input to the prediction model is obtained by concatenating the deposit and ore sub-features, the amount of data in the deposit and ore sub-features is smaller than that in the deposit and ore features, resulting in lower computational overhead and higher processing efficiency for the prediction model.

[0058] In related technologies, increasing the amount of input data to the prediction model generally improves the accuracy of the output target mineral deposit reserves. However, mineral deposit and ore information are high-dimensional and exhibit significant grouping attributes, leading to an extremely sparse feature space composed of mineral deposit and ore features. This increases the spurious correlation between noise and key features, making it difficult for the prediction model to capture effective features for prediction. After segmentation, the dimensionality of the sub-features decreases, and the sample density increases significantly, which is beneficial for the prediction model to learn the relationship between the first concatenated input features and the output target mineral deposit reserves.

[0059] According to an embodiment of the present invention, the deposit features and ore features are first segmented to obtain multiple deposit sub-features and multiple ore sub-features. Then, one sub-feature is randomly selected from each of these deposit sub-features and ore sub-features and spliced ​​together to obtain a first spliced ​​feature. The target deposit reserve is determined using the first spliced ​​feature, which reduces the probability of overfitting of the prediction model due to excessive feature data input to the prediction model, thereby improving the accuracy of the target deposit reserve. The first spliced ​​feature obtained by splicing the randomly selected deposit sub-features and ore sub-features can construct explicit feature interaction, helping the prediction model to capture features with high correlation to the predicted target deposit reserve in the first spliced ​​feature, further improving the accuracy of the target deposit reserve.

[0060] According to an embodiment of the present invention, the ore reserve prediction method further includes: extracting data from the ore deposit to be predicted based on the ore deposit dataset to obtain an initial ore deposit name; and standardizing the initial ore deposit name according to a preset naming template to obtain the ore deposit name.

[0061] According to an embodiment of the present invention, the mineral deposit dataset includes information about the mineral deposit to be predicted, and may include an initial name for the mineral deposit to be predicted. Data is extracted from the mineral deposit to be predicted in the mineral deposit dataset to obtain an initial name, for example, the initial name may be "Mineral Deposit No. 1 in a certain province" or "Mineral Deposit No. 1 located in a certain province".

[0062] The deposit name is obtained by extracting elements corresponding to a preset naming template from the initial deposit name. The preset naming template includes the province and the deposit number. For example, the province element "A" and the deposit number element "1" are extracted from "Deposit No. 1 in Province A". The extracted elements are then used to fill the naming template "Deposit No. * in Province *", resulting in the deposit name "Deposit No. 1 in Province A".

[0063] According to embodiments of the present invention, by standardizing the names of mineral deposits, the impact of inconsistent mineral deposit naming standards on prediction models is reduced, thereby improving the accuracy of target mineral deposit reserves.

[0064] According to an embodiment of the present invention, the deposit information includes the deposit name and deposit thickness, and the ore information includes the sampling depth and the ore density corresponding to the sampling depth.

[0065] According to embodiments of the present invention, the name of a mineral deposit may include a geographical location and number representing the deposit to be predicted. By using the deposit name as one piece of information for prediction, the prediction model can capture the geographical location information of the deposit to be predicted based on the features corresponding to the deposit name. Since deposits in the same geographical location tend to have similar formation reasons and similar ore distribution patterns, the geographical location information captured by the prediction model from the deposit name can assist in predicting the target deposit's reserves. The deposit thickness can be obtained using mineral deposit surveying equipment.

[0066] It is understandable that in the same region, mineral deposits have similar mineralization causes, and therefore, the distribution of ore within them is similar. Furthermore, most mineral deposits exhibit a layered structure, and ore densities at the same depth are similar. Therefore, during the exploration of mineral deposits, multiple sampling depths can be selected to measure the ore density at different depths.

[0067] The name of the deposit contains information that characterizes the regional geological patterns, which makes up for the lack of sampling points. The deposit thickness provides the vertical scale of the deposit, which is used to determine the spatial scale of the ore body. The sampling depth and the ore density corresponding to the sampling depth can provide ore density information in the vertical direction for the prediction model.

[0068] According to an embodiment of the present invention, based on the geological laws of ore deposit formation, the ore deposit name, ore deposit thickness, sampling depth, and ore density corresponding to the sampling depth are used as information for determining the target ore deposit reserves. In order to control the amount of input data for the prediction model, the possibility of overfitting of the prediction model is reduced and the accuracy of the target ore deposit reserves is improved.

[0069] According to an embodiment of the present invention, the deposit features include deposit name features and deposit thickness features, and the ore features include sampling depth features and ore density features. The deposit features and ore features are segmented according to a preset feature segmentation dimension to obtain multiple deposit sub-features and multiple ore sub-features, including: horizontally splicing the sampling depth features and ore density features to obtain a third spliced ​​feature; and segmenting the deposit name features, deposit thickness features, and third spliced ​​feature according to a preset feature segmentation dimension to obtain multiple name sub-features, multiple thickness sub-features, and multiple ore sub-features.

[0070] According to an embodiment of the present invention, the sampling depth feature and the ore density feature are horizontally concatenated to bind the correspondence between the feature sequences of the sampling depth feature and the ore depth feature, thereby obtaining a third concatenated feature. Horizontal concatenation refers to merging the sampling depth feature and the ore density feature in a column direction. This ensures that after feature segmentation of the deposit name feature, deposit thickness feature, and third concatenated feature according to a preset feature segmentation dimension, the feature sequences of the sampling depth feature and the ore depth feature can still correspond to each other.

[0071] For example, sampling depth features: [sampling depth sub-feature 1, sampling depth sub-feature 2] T Ore depth characteristics: [Ore density sub-feature 1, Ore density sub-feature 2] T Where T represents the transpose of the matrix, the sampling depth feature and the ore density feature are horizontally concatenated to obtain the third concatenated feature: {[Sampling depth sub-feature 1, Sampling depth sub-feature 2]} T[Ore density sub-feature 1, ore density sub-feature 2] T Since the third stitching feature is obtained by horizontally stitching the sampling depth feature and the ore density feature, after feature segmentation of the third stitching feature, it can still be guaranteed that the sampling depth sub-feature 1 corresponds to the ore depth sub-feature 1.

[0072] According to an embodiment of the present invention, the sampling depth feature and the ore density feature are horizontally spliced ​​to obtain a third spliced ​​feature before feature segmentation is performed. This can avoid disrupting the feature sequence correspondence between the sampling depth feature and the ore density feature during the subsequent random selection of sub-features, thereby reducing the probability of the prediction model learning incorrect information and improving the accuracy of the target ore deposit reserves.

[0073] According to an embodiment of the present invention, one sub-feature is randomly selected from multiple name sub-features, multiple thickness sub-features, and multiple ore sub-features for splicing to obtain a fourth spliced ​​feature for input into the prediction model; the fourth spliced ​​feature is input into the prediction model to output the target ore deposit reserves of the ore deposit to be predicted, so as to reduce the correlation between non-critical features before the features are input into the prediction model and improve the accuracy of the target ore deposit reserves.

[0074] Figure 3 A schematic diagram of feature segmentation according to an embodiment of the present invention is shown.

[0075] like Figure 3 As shown, the deposit name feature 301, deposit thickness feature 302, and third splicing feature 303 are segmented according to the preset feature segmentation dimensions to obtain multiple deposit name sub-features, multiple deposit thickness sub-features, and multiple third splicing sub-features. A deposit name sub-feature 304, a deposit thickness sub-feature 305, and a third splicing sub-feature 306 are randomly selected. The deposit name sub-feature 304, the deposit thickness sub-feature 305, and the third splicing feature 306 are spliced ​​to obtain feature 307 used to predict the target deposit reserves of the deposit to be predicted.

[0076] The ore deposit name feature 301 and ore deposit thickness feature 302 are segmented respectively to randomly select ore deposit name sub-feature 304, ore deposit thickness sub-feature 305 and a third splicing sub-feature 306 to determine the feature 307 for predicting the target ore deposit reserves. This reduces the probability of the prediction model overfitting due to the large amount of feature data input to the prediction model and improves the accuracy of the target ore deposit reserves.

[0077] According to an embodiment of the present invention, the prediction model includes a self-attention module and an evaluation module; inputting a first splicing feature into the prediction model and outputting the target mineral deposit reserve of the mineral deposit to be predicted includes: inputting the first splicing feature into the self-attention module and outputting an interaction feature; inputting the interaction feature into the evaluation module so that the evaluation module evaluates multiple candidate mineral deposit reserves based on the interaction feature and outputs the target mineral deposit reserve, wherein the multiple candidate mineral deposit reserves are determined during the training of the prediction model, and the multiple candidate mineral deposit reserves can be the labels of all training samples used in the model training.

[0078] According to an embodiment of the present invention, a self-attention module is used to capture features at different levels in the first stitched features. The first stitched features are input into the self-attention module to facilitate the extraction of interaction relationships between features for feature enhancement, and to output interactive features. These interactive features characterize the interaction between deposit information and ore information.

[0079] The evaluation module assesses the reserves of multiple candidate mineral deposits and determines the target mineral deposit with the highest confidence level. Interactive features are input into the evaluation module, which then evaluates the reserves of multiple candidate mineral deposits based on these features to identify the candidate mineral deposit with the highest confidence level corresponding to the interactive features, which is then used as the target mineral deposit.

[0080] According to an embodiment of the present invention, a self-attention module is used to determine the interaction features for evaluation, so as to improve the feature representation ability of the interaction features, make up for the problem of reduced learnable features caused by feature segmentation, and improve the accuracy of the target mineral deposit reserves.

[0081] According to an embodiment of the present invention, the self-attention module includes multiple attention sub-modules; inputting a first concatenated feature into the self-attention module and outputting an interactive feature includes: inputting the first concatenated feature into multiple attention sub-modules respectively and outputting multiple interactive sub-features; concatenating the multiple interactive sub-features to obtain a second concatenated feature; and reconstructing the second concatenated feature to obtain the interactive feature.

[0082] According to an embodiment of the present invention, the second concatenated feature is input into multiple attention sub-modules, so that the multiple attention sub-modules can extract the interaction relationship between features based on the weights determined during training, and so that the multiple attention sub-modules can output their respective interaction sub-features. Each attention sub-module focuses on different parts of the input second concatenated feature.

[0083] Multiple interactive sub-features are sequentially concatenated to form a larger feature matrix, preserving the diverse information captured by each attention sub-module, resulting in the second concatenated feature.

[0084] The second concatenated feature is reconstructed into a more expressive feature, thus obtaining the interactive feature.

[0085] According to an embodiment of the present invention, the first splicing feature is processed by multiple attention submodules to obtain interactive features, thereby improving the expressive power of the interactive features and thus improving the accuracy of the target mineral deposit reserves.

[0086] According to an embodiment of the present invention, feature reconstruction of the second splicing feature to obtain interactive features includes: performing a linear transformation on the second splicing feature according to a linear transformation function to obtain transformed features; and performing nonlinear reconstruction on the transformed features according to a nonlinear transformation function to obtain interactive features.

[0087] According to an embodiment of the present invention, the transformation features can be determined by formula (2):

[0088] Formula (2);

[0089] in, Characterizing linear transformation functions, As variables, , The number of attention submodules, Characterizing the first One interactive sub-feature, This is a learnable weight matrix. It can be expressed by formula (3):

[0090] Formula (3);

[0091] in, , , , This is the second splicing feature. Let be the query matrix used by the i-th attention submodule to project the second concatenated feature; Let be the key matrix used by the i-th attention submodule to project the second concatenated feature; Let be the value matrix used by the i-th attention submodule to project the second concatenated feature; This is the transpose of the key matrix; The dimensions of the query matrix and the key matrix.

[0092] Nonlinear transformation function It can be expressed by formula (4):

[0093] Formula (4);

[0094] in, , To map transform features to a learnable parameter matrix in the semantic space, , These are learnable parameters. This is the activation function.

[0095] According to an embodiment of the present invention, a linear transformation is performed on the second splicing feature according to a linear transformation function to obtain a transformed feature, thereby transforming the matrix dimension of the second splicing feature; the transformed feature is then nonlinearly reconstructed to nonlinearly enhance the transformed feature, resulting in an interactive feature containing rich contextual information, which improves the expressive power of the interactive feature and thus improves the accuracy of the target mineral deposit reserves.

[0096] According to an embodiment of the present invention, inputting interactive features into an evaluation module so that the evaluation module evaluates the reserves of multiple candidate mineral deposits based on the interactive features and outputs the target mineral deposit reserves includes: when the evaluation module receives the interactive features, performing a nonlinear transformation on the interactive features to enhance the semantic expressive power of the interactive features and obtain enhanced feature values; determining the target mineral deposit reserves among the multiple candidate mineral deposit reserves based on the matching degree between the enhanced feature values ​​and the candidate reserve quantum features of the multiple candidate mineral deposit reserves, and outputting the target mineral deposit reserves.

[0097] According to an embodiment of the present invention, when the evaluation module receives the interaction features, a feedforward neural network can be used to perform a nonlinear transformation on the interaction features based on the sampling depth features to enhance the semantic expressive power of the interaction features and obtain enhanced feature values. The nonlinear transformation on the interaction features based on the sampling depth features can be expressed by formula (5):

[0098] Formula (5);

[0099] in, For sampling depth sub-features, This is the matrix transpose of the sampling depth sub-features. , For interactive features, To enhance the eigenvalues.

[0100] Candidate reserve quantum features are features used to train the prediction model during the training process. They can be obtained by feature segmentation of features corresponding to candidate deposit reserves. The evaluation results of multiple candidate deposit reserves are calculated based on the enhancement feature values ​​to determine the matching degree between the multiple candidate reserve quantum features and the enhancement feature values. The target deposit reserve is then determined from among the multiple candidate deposit reserves, and the target deposit reserve is output.

[0101] According to an embodiment of the present invention, a nonlinear transformation is performed on the interaction features based on the sampling depth features to determine the enhanced feature value that characterizes the semantic interaction between the sampling depth features and the interaction features. The target mineral deposit reserves are then determined based on the enhanced feature value, thereby improving the accuracy of the target mineral deposit reserves.

[0102] According to an embodiment of the present invention, determining a target mineral deposit reserve among multiple candidate mineral deposit reserves based on the matching degree between the enhanced feature value and the candidate quantum characteristics of each of the multiple candidate mineral deposit reserves includes: for each candidate mineral deposit reserve, determining a target attribute feature value with the same sampling depth as the ore information from multiple initial attribute feature values, wherein the initial attribute feature value is the product of the mineral deposit sample features and the ore sample features used during training; determining a potential feature value based on the product of the mineral deposit sub-features, ore sub-features, candidate quantum characteristics, and ore density gradient; weighting and summing the potential feature values, target attribute feature values, and enhanced feature values ​​of each training round based on multiple weight parameters obtained in each training round to obtain a first evaluation value; determining a second evaluation value and the first evaluation value based on the product of the mineral deposit sub-features, ore sub-features, and candidate quantum characteristics to determine the matching degree between the enhanced feature value and the candidate quantum characteristics of the candidate mineral deposit reserve; and determining the target mineral deposit reserve among multiple candidate mineral deposit reserves based on the matching degree of each candidate mineral deposit reserve.

[0103] According to an embodiment of the present invention, the initial attribute feature value is the product of the deposit sample features and the ore sample features used during the training process. The deposit sample features are the features corresponding to the deposit information used during the training process, and the ore sample features are the features corresponding to the ore information used during the training process. The data used during the training process comes from multiple deposits, each with a different sampling depth, which can provide multiple training samples for the training process.

[0104] Ore information corresponds to sampling depth. Among multiple initial attribute feature values, the target attribute feature value is selected based on the sampling depth corresponding to the ore information. This target attribute feature value is used to calculate the matching degree between the enhancement feature value and the candidate reservoir quantum features of each of the multiple candidate ore deposits. The ore density gradient is used to characterize the rate of change of the ore density of the ore deposit to be predicted with sampling depth.

[0105] Among multiple candidate mineral deposit reserves, a second evaluation value and a first evaluation value are determined for each candidate mineral deposit reserve. The candidate mineral deposit reserve with the largest sum of the first evaluation value and the second evaluation value is determined as the target mineral deposit reserve.

[0106] The sum of the first and second assessment values ​​can be expressed by formula (6):

[0107] Formula (6);

[0108] in, Transpose that characterizes the sub-features of a mineral deposit. Characterizing the features of ore particles Characterizing candidate storage quantum features, Characterizing sampling depth sub-features, Characterizes the first evaluation value. Indicates the total number of training rounds. , and This represents the learnable weight parameter matrix at the sampling depth of the sampling depth sub-feature representation in the k-th training round.

[0109] Characterizes the second evaluation value. Characterize latent eigenvalues, , Characterizing the density gradient value of the ore, ,in, It is a natural constant. These are preset parameters used to control ore density. With sampling depth Sensitivity to change The smaller the value, the lower the ore density. With sampling depth The less obvious the change.

[0110] Characterizing the feature values ​​of the target attribute, where, The matrix representing the features of the ore deposit samples using the training samples is defined by the sampling depth sub-feature characterization. Characterization pairs Flattening, soon Convert to a vector; The matrix representing the ore sample features using training samples is defined by the sampling depth sub-feature representation. Characterization pairs Flattening, soon Convert to a vector. Characterize the enhanced eigenvalues.

[0111] According to an embodiment of the present invention, a first evaluation value is used as a base item to provide a static benchmark evaluation for mineral deposit reserve prediction, and a second evaluation value is used to correct the total result to help the prediction model determine the highest target mineral deposit reserve by the sum of the first evaluation value and the second evaluation value.

[0112] Figure 4 A schematic diagram of a mineral deposit reserve prediction method according to an embodiment of the present invention is shown.

[0113] like Figure 4As shown, the first splicing feature 401 is input into the attention module M410, which outputs the interaction feature 402. The evaluation module M420 performs a nonlinear transformation on the interaction feature 402 to enhance its semantic expressive power, resulting in an enhanced feature value 403. Based on the enhanced feature value 403, the matching degree of the candidate reservoir quantum features 404 for each of the multiple candidate ore deposit reserves is calculated. Specifically, for each candidate ore deposit reserve, a target attribute feature value 406 with the same sampling depth as the ore information is determined from multiple initial attribute feature values ​​405. Based on the ore deposit sub-features 407, ore sub-features 408, candidate reservoir quantum features 404, and ore density gradient 409... The product of the potential feature value 410 is determined; based on the multiple weight parameters obtained in each training round, the potential feature value 410, the target attribute feature value 406, and the reinforcement feature value 403 of each training round are weighted and summed to obtain the first evaluation value 411; the second evaluation value 412 is determined according to the product of the ore sub-feature 407, the ore sub-feature 408 and the candidate reservoir quantum feature 404; the matching degree 413 between the candidate ore deposit reserve corresponding to the candidate reservoir quantum feature 404 and the reinforcement feature value 403 is determined according to the matching degree 413; the target ore deposit reserve 414 is determined and output according to the matching degree 413 among multiple candidate ore deposit reserves.

[0114] Figure 5 A flowchart illustrating a training method for a prediction model according to an embodiment of the present invention is shown.

[0115] like Figure 5 As shown, the training method of the prediction model in this embodiment includes operations S510 to S550.

[0116] In operation S510, multiple sample sets are randomly determined from the sample sets of multiple ore deposits. Each sample set includes multiple sample data of the same ore deposit, and the sample data includes sample ore deposit information, sample ore information, and sample reserve information.

[0117] In operation S520, feature encoding is performed on the sample deposit information, sample ore information, and sample reserve information to obtain sample deposit characteristics, sample ore characteristics, and sample reserve characteristics.

[0118] In operation S530, feature segmentation is performed on the sample deposit features and sample ore features according to the preset feature segmentation dimensions to obtain multiple sample deposit sub-features and multiple sample ore sub-features.

[0119] In operation S540, one sub-feature is randomly selected from multiple sample deposit sub-features and multiple sample ore sub-features for feature splicing to obtain the first sample spliced ​​feature.

[0120] In operation S550, the first sample splicing features are input into the initial prediction model, and the predicted reserve features are output. The loss value of the initial prediction model is calculated based on the sample reserve features and the predicted reserve features. This operation is repeated until the loss value meets the preset conditions, thus obtaining the prediction model. According to embodiments of the present invention, descriptions of sample deposit information, sample ore information, etc., can be found in the above-mentioned content regarding deposit information and ore information, and will not be repeated here.

[0121] To improve the accuracy of the prediction model, multiple first splicing features can be determined by using all permutations and combinations of multiple sample deposit sub-features and multiple sample ore sub-features, which can then be used to train the initial prediction model.

[0122] The initial prediction model is used to calculate the first sample splicing features of the input, so as to determine the predicted reserve characteristics of the deposit based on the information represented in the first sample splicing features. The loss value of the initial prediction model is calculated according to the sample reserve characteristics and the predicted reserve characteristics. The loss value can be determined by formula (7):

[0123] Formula (7);

[0124] in, Multiple sample sets used for representation training It is a constant. The set represents the individual sample storage features of multiple sample data used for training; the inner summation sign represents the representation; calculation... The storage features of each sample are derived from the set of storage features of each of the multiple sample data; the summation symbol in the outer layer represents the summation of the loss values ​​of the multiple sample data used for training, resulting in the final total loss value.

[0125] Characterizing the predicted reserves, Characterizes the sample's storage characteristics. A function is an exponential function in mathematics with the natural constant e as its base.

[0126] The initial prediction model is trained repeatedly until the loss value meets the preset conditions, thus obtaining the prediction model.

[0127] According to an embodiment of the present invention, the sample deposit features and sample ore features are first segmented to obtain multiple sample deposit sub-features and multiple sample ore sub-features. Then, a sub-feature is randomly selected from each of these sample deposit sub-features and sample ore sub-features and concatenated to obtain a first sample concatenated feature. The first sample concatenated feature is used as input to the initial prediction model to adjust the parameters in the initial prediction model based on the loss function. This reduces the probability of overfitting of the prediction model due to excessive feature data input to the prediction model and improves the accuracy of the target deposit reserves. The first sample concatenated feature obtained by concatenating the randomly selected sample deposit sub-features and sample ore sub-features can construct an explicit feature interaction, helping the initial prediction model to capture features with high correlation to the predicted reserve features in the first sample concatenated feature, further improving the accuracy of the target deposit reserves.

[0128] Based on the above-mentioned mineral deposit reserve prediction method, this invention also provides a mineral deposit reserve prediction device. The following will be combined with... Figure 6 The device is described in detail.

[0129] Figure 6 A structural block diagram of a mineral deposit reserve prediction device according to an embodiment of the present invention is shown.

[0130] like Figure 6 As shown, the mineral deposit reserve prediction device 600 of this embodiment includes a first encoding module 610, a first segmentation module 620, a first splicing module 630 and a first prediction module 640.

[0131] The first encoding module 610 is used to perform feature encoding on the ore deposit information and the ore information of the ore deposit to be predicted, respectively, to obtain ore deposit features and ore features. In one embodiment, the encoding module 610 can be used to perform the operation S210 described above, which will not be repeated here.

[0132] The first segmentation module 620 is used to segment the deposit features and ore features according to a preset feature segmentation dimension, respectively, to obtain multiple deposit sub-features and multiple ore sub-features. In one embodiment, the segmentation module 620 can be used to perform the operation S220 described above, which will not be repeated here.

[0133] The first splicing module 630 is used to randomly select one sub-feature from multiple deposit sub-features and multiple ore sub-features for splicing to obtain a first spliced ​​feature. In one embodiment, the splicing module 630 can be used to perform the operation S230 described above, which will not be repeated here.

[0134] The first prediction module 640 is used to input the first splicing feature into the prediction model and output the target mineral deposit reserves of the deposit to be predicted. In one embodiment, the prediction module 640 can be used to perform the operation S240 described above, which will not be repeated here.

[0135] According to embodiments of the present invention, any plurality of modules among the first encoding module 610, the first segmentation module 620, the first splicing module 630, and the first prediction module 640 may be combined into one module, or any one of these modules may be split into multiple modules. Alternatively, at least a portion of the functionality of one or more of these modules may be combined with at least a portion of the functionality of other modules and implemented in one module. According to embodiments of the present invention, at least one of the first encoding module 610, the first segmentation module 620, the first splicing module 630, and the first prediction module 640 may be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging the circuitry, or implemented in software, hardware, or firmware, or in any one of the three implementation methods, or in a suitable combination of any of them. Alternatively, at least one of the first encoding module 610, the first segmentation module 620, the first splicing module 630, and the first prediction module 640 may be implemented at least partially as a computer program module, which can perform corresponding functions when the computer program module is run.

[0136] Based on the above-described training method for the prediction model, this invention also provides a training apparatus for the prediction model. The following will combine... Figure 7 The device is described in detail.

[0137] Figure 7 A structural block diagram of a training apparatus for a prediction model according to an embodiment of the present invention is shown.

[0138] like Figure 7 As shown, the mineral deposit reserve prediction device 700 of this embodiment includes a set determination module 710, a second encoding module 720, a second segmentation module 730, a second splicing module 740, and a second prediction module 750.

[0139] The set determination module 710 is used to randomly determine multiple sample sets from the sample sets of multiple mineral deposits. The sample sets include multiple sample data of the same mineral deposit, and the sample data includes sample mineral deposit information, sample ore information and sample reserve information.

[0140] The second encoding module 720 is used to perform feature encoding on the sample deposit information, sample ore information and sample reserve information respectively to obtain sample deposit features, sample ore features and sample reserve features.

[0141] The second segmentation module 730 is used to perform feature segmentation on the sample deposit features and sample ore features according to the preset feature segmentation dimensions, so as to obtain multiple sample deposit sub-features and multiple sample ore sub-features.

[0142] The second splicing module 740 is used to randomly select one sub-feature from multiple sample deposit sub-features and multiple sample ore sub-features for feature splicing to obtain the first sample spliced ​​feature.

[0143] The second prediction module 750 is used to input the spliced ​​features of the first sample into the initial prediction model and output the predicted reserve features; calculate the loss value of the initial prediction model based on the sample reserve features and the predicted reserve features; repeat the above operations until the loss value meets the preset conditions to obtain the prediction model.

[0144] According to embodiments of the present invention, any plurality of modules among the set determination module 710, the second encoding module 720, the second segmentation module 730, the second splicing module 740, and the second prediction module 750 can be merged into one module, or any one of these modules can be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules can be combined with at least part of the functionality of other modules and implemented in one module. According to embodiments of the present invention, at least one of the set determination module 710, the second encoding module 720, the second segmentation module 730, the second splicing module 740, and the second prediction module 750 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging circuitry, or implemented in hardware or firmware, or in any one of software, hardware, and firmware implementations, or in a suitable combination of any of these. Alternatively, at least one of the set determination module 710, the second encoding module 720, the second segmentation module 730, the second splicing module 740, and the second prediction module 750 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0145] Figure 8 A block diagram of an electronic device suitable for implementing a mineral deposit reserve prediction method according to an embodiment of the present invention is shown.

[0146] like Figure 8As shown, an electronic device 800 according to an embodiment of the present invention includes a processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage portion 808 into a random access memory (RAM) 803. The processor 801 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 801 may also include onboard memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0147] RAM 803 stores various programs and data required for the operation of electronic device 800. Processor 801, ROM 802, and RAM 803 are interconnected via bus 804. Processor 801 executes various operations of the method flow according to embodiments of the present invention by executing programs in ROM 802 and / or RAM 803. It should be noted that the programs may also be stored in one or more memories other than ROM 802 and RAM 803. Processor 801 may also execute various operations of the method flow according to embodiments of the present invention by executing programs stored in said one or more memories.

[0148] According to an embodiment of the present invention, the electronic device 800 may further include an input / output (I / O) interface 805, which is also connected to a bus 804. The electronic device 800 may also include one or more of the following components connected to the input / output (I / O) interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the input / output (I / O) interface 805 as needed. A removable medium 811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 810 as needed so that computer programs read from it can be installed into the storage section 808 as needed.

[0149] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of the present invention.

[0150] According to embodiments of the present invention, a computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of the present invention, a computer-readable storage medium may include ROM 802 and / or RAM 803 and / or one or more memories other than ROM 802 and RAM 803 described above.

[0151] Embodiments of the present invention also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to enable the computer system to implement the mineral deposit reserve prediction method provided in the embodiments of the present invention.

[0152] When the computer program is executed by the processor 801, it performs the functions defined in the system / apparatus of this invention. According to embodiments of the invention, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0153] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 809, and / or installed from a removable medium 811. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0154] In such an embodiment, the computer program can be downloaded and installed from a network via communication section 809, and / or installed from removable medium 811. When the computer program is executed by processor 801, it performs the functions defined in the system of this embodiment of the invention. According to embodiments of the invention, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0155] According to embodiments of the present invention, program code for executing the computer programs provided in the embodiments of the present invention can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0156] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0157] Those skilled in the art will understand that the features described in the various embodiments of the present invention can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, the features described in the various embodiments of the present invention can be combined and / or combined in various ways without departing from the spirit and teachings of the present invention. All such combinations and / or combinations fall within the scope of the present invention.

[0158] The embodiments of the present invention have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of the invention. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of the invention, and all such substitutions and modifications should fall within the scope of the invention.

Claims

1. A method of ore deposit reserve estimation, characterized by, The method comprises: respectively feature coding deposit information of a to-be-predicted deposit and ore information for the to-be-predicted deposit, to obtain deposit features and ore features, wherein the deposit features comprise deposit name features and deposit thickness features, and the ore features comprise sampling depth features and ore density features; transversely splicing the sampling depth features and the ore density features to bind the correspondence between the sampling depth features and the ore density features on a feature sequence, to obtain third spliced features; respectively feature segmenting the deposit name features, the deposit thickness features and the third spliced features according to a preset feature segmentation dimension, so that each sub-feature contains multiple dimension information, the correlation between each of the sub-features is low, and the correlation within each of the sub-features is high, to obtain multiple name sub-features, multiple thickness sub-features and multiple ore sub-features; randomly determining one sub-feature from the multiple name sub-features, the multiple thickness sub-features and the multiple ore sub-features respectively for splicing, to obtain first spliced features; inputting the first spliced features into a prediction model, and outputting a target deposit reserve of the to-be-predicted deposit.

2. The method of claim 1, wherein, The prediction model comprises a self-attention module and an evaluation module; the inputting the first spliced features into the prediction model and the outputting the target deposit reserve of the to-be-predicted deposit comprise: inputting the first spliced features into the self-attention module, and outputting interaction features; inputting the interaction features into the evaluation module, so that the evaluation module evaluates multiple candidate deposit reserves based on the interaction features, and outputs the target deposit reserve, wherein the multiple candidate deposit reserves are determined in a process of training the prediction model.

3. The method of claim 2, wherein, The self-attention module comprises multiple attention sub-modules; the inputting the first spliced features into the self-attention module and the outputting interaction features comprise: inputting the first spliced features into the multiple attention sub-modules respectively, and outputting multiple interaction sub-features; feature splicing the multiple interaction sub-features, to obtain second spliced features; feature reconstructing the second spliced features, to obtain the interaction features.

4. The method of claim 3, wherein, The feature reconstructing the second spliced features, to obtain the interaction features, comprises: linearly transforming the second spliced features according to a linear transformation function, to obtain transformed features; nonlinearly reconstructing the transformed features according to a nonlinear transformation function, to obtain the interaction features.

5. The method of claim 2, wherein, The inputting the interaction features into the evaluation module, so that the evaluation module evaluates multiple candidate deposit reserves based on the interaction features and outputs the target deposit reserve, comprises: in a case where the evaluation module receives the interaction features, nonlinearly transforming the interaction features, to strengthen semantic expression capability of the interaction features, to obtain strengthened feature values; determining the target deposit reserve from the multiple candidate deposit reserves according to matching degrees between the strengthened feature values and candidate reserve sub-features of the multiple candidate deposit reserves respectively, to output the target deposit reserve.

6. The method of claim 5, wherein, The method comprises the following steps: For each candidate ore deposit reserve, Determine a target attribute feature value corresponding to the sampling depth in a plurality of initial attribute feature values, wherein the initial attribute feature value is the product of the deposit sample feature and the ore sample feature used in the training process; Determine a potential feature value according to the product of the name sub-feature, the thickness sub-feature, the ore sub-feature, the candidate reserve sub-feature and the ore density gradient; Based on the plurality of weight parameters obtained in each training round, the potential feature value, the target attribute feature value and the reinforced feature value of each training round are weighted and summed to obtain a first evaluation value; Determine the matching degree of the reinforced feature value and the candidate reserve sub-feature of the candidate ore deposit reserve according to the second evaluation value determined by the product of the name sub-feature, the thickness sub-feature, the ore sub-feature and the candidate reserve sub-feature and the first evaluation value; Determine the target ore deposit reserve from a plurality of candidate ore deposit reserves according to the matching degree of each candidate ore deposit reserve.

7. The method according to any one of claims 1 to 6, characterized in that, The deposit information includes deposit name and deposit thickness, and the ore information includes sampling depth and ore density corresponding to the sampling depth.

8. The method of claim 7, wherein, The method further comprises: Based on the deposit data set, data extraction is performed on the to-be-predicted deposit to obtain an initial deposit name; According to a preset naming template, the initial deposit name is standardized to obtain the deposit name. 9.A method for training a prediction model, the method comprising: The method comprises: Randomly determine a plurality of sample sets in a plurality of sample sets of each deposit, wherein the sample set comprises a plurality of sample data of the same deposit, and the sample data comprises sample deposit information, sample ore information and sample reserve information; Feature coding is performed on the sample deposit information, the sample ore information and the sample reserve information respectively to obtain sample deposit features, sample ore features and sample reserve features, wherein the sample deposit features comprise sample deposit name features and sample deposit thickness features, and the sample ore features comprise sample sampling depth features and sample ore density features; The sample sampling depth features and the sample ore density features are transversely spliced to bind the corresponding relationship between the sample sampling depth features and the sample ore density features in the feature sequence to obtain a sample third spliced feature; According to a preset feature segmentation dimension, feature segmentation is performed on the sample deposit name features, the sample deposit thickness features and the sample third spliced features respectively, so that each sub-feature contains a plurality of dimension information, the correlation between each sub-feature is low, and the correlation within each sub-feature is high, to obtain a plurality of sample name sub-features, a plurality of sample thickness sub-features and a plurality of sample ore sub-features; Randomly determine one sub-feature in the plurality of sample name sub-features, the plurality of sample thickness sub-features and the plurality of sample ore sub-features for feature splicing to obtain a first sample spliced feature; inputting the first sample splicing feature into an initial prediction model to output a predicted reserve feature; calculating a loss value of the initial prediction model according to the sample reserve feature and the predicted reserve feature; repeating the above operations until the loss value meets a preset condition to obtain a prediction model.

Images