Multi-path fusion intelligent prospecting prediction model construction method, device and equipment

Abstract

The application provides a multi-path fusion intelligent ore-prospecting prediction model construction method, device and equipment, effectively solving the problem that the existing manually created ore-prospecting prediction model cannot meet the requirements of intelligentization, quantification and interpretability. The method comprises the following steps: acquiring multi-source heterogeneous data related to intelligent ore prospecting of a to-be-predicted area, and preprocessing the multi-source heterogeneous data according to the data types of the multi-source heterogeneous data to obtain target multi-source heterogeneous data; calling a pre-configured multi-path in parallel to mine ore-prospecting prediction elements in the target multi-source heterogeneous data in parallel, and obtaining element sets corresponding to the multi-path respectively; fusing the element sets corresponding to the multi-path respectively to obtain a full prediction element system, and constructing an intelligent ore-prospecting prediction model based on the full prediction element system; predicting a deposit type of the to-be-predicted area based on the intelligent ore-prospecting prediction model to obtain a prediction result, and continuously optimizing the intelligent ore-prospecting prediction model based on the prediction result and new survey results.

Description

Technical Field

[0001] This application relates to the field of model building technology, and more specifically, to a method, apparatus, and equipment for building intelligent mineral exploration prediction models using multi-pathway fusion. Background Technology

[0002] Mineral resource exploration is characterized by high investment, high risk, and complexity, making the construction of a scientific and reliable mineral exploration prediction model crucial. This model can systematically integrate massive amounts of heterogeneous exploration data and geological knowledge, transforming qualitative geological understanding into calculable and spatially quantifiable predictive elements. This enables the quantitative evaluation and selection of mineralized potential areas, meeting the core needs of modern mineral exploration.

[0003] Currently, mineral exploration prediction model construction is primarily driven by expert knowledge, relying heavily on the experience and knowledge of geological experts. It combines comprehensive analysis and manual modeling with multi-source geological exploration data from limited areas to ultimately form a prediction model that primarily uses qualitative description and semi-quantitative spatial overlay. The specific process is as follows: Figure 1 As shown, various exploration data, including geological, geophysical, geochemical, and remote sensing data, are collected. Experts conduct comprehensive analysis based on the theory of typical mineral deposit metallogenic systems, manually interpret and extract mineral exploration indicators such as structure and lithology, and then systematically organize them into a conceptual model. These are then abstracted into predictive factors and assigned weights manually to complete the prediction.

[0004] Existing technologies have significant limitations: First, the models are highly subjective, with core knowledge fixed in expert experience or simple configuration tables, lacking unified quantitative and computational expression, resulting in poor model repeatability. Second, conceptual models are fragmented and isolated, with experts limiting themselves to single mineral types, single regions, or single stages of analysis to simplify the models, making it impossible to achieve correlational reasoning at the levels of geological bodies and mineralization, and limiting their interpretability and transferability. Third, when dealing with complex metallogenic systems and massive amounts of data, they lack unified semantic expression and statistical modeling capabilities, making it difficult to handle high-dimensional and nonlinear features, unable to leverage the advantages of modern information technology in correlation mining and probabilistic reasoning, and failing to meet the requirements of mineral exploration breakthroughs for prediction accuracy, efficiency, and objectivity.

[0005] Therefore, developing more intelligent, quantitative, and interpretable mineral exploration prediction models has become an inevitable trend and core requirement for the development of modern mineral exploration technology. Summary of the Invention

[0006] In view of this, the purpose of this application is to provide a method, apparatus and equipment for constructing a multi-pathway integrated intelligent mineral exploration prediction model, which effectively solves the problem that existing manually created mineral exploration prediction models cannot meet the requirements of intelligence, quantification and interpretability.

[0007] In a first aspect, embodiments of this application provide a method for constructing a multi-pathway fusion intelligent mineral exploration prediction model, the method comprising: Acquire multi-source heterogeneous data related to intelligent mineral exploration in the area to be predicted, and preprocess the multi-source heterogeneous data according to the data type of the multi-source heterogeneous data to obtain target multi-source heterogeneous data; The system invokes pre-configured multiple approaches in parallel to mine mineral exploration prediction elements in the target multi-source heterogeneous data, obtaining element sets corresponding to each approach; the multiple approaches include at least model mining, knowledge mining, and spatial mining. A comprehensive prediction element system is obtained by integrating the element sets corresponding to multiple approaches, and an intelligent mineral exploration prediction model is constructed based on the comprehensive prediction element system. Based on the intelligent mineral exploration prediction model, interpretable prediction results are obtained according to the type of mineral deposit in the area to be predicted. The intelligent mineral exploration prediction model is then continuously optimized based on the prediction results and newly added exploration results.

[0008] In conjunction with the first aspect, this application provides a first possible implementation of the first aspect, wherein the pre-configured multiple paths are implemented based on the following steps: The multiple pathways are pre-configured so that each pathway has a corresponding processing method; different pathways correspond to different processing methods. The processing method is executed in parallel to mine mineral exploration prediction elements in the target multi-source heterogeneous data.

[0009] In conjunction with the first aspect, this application provides a second possible implementation of the first aspect, wherein the multiple approaches include at least model mining; Mining mineral exploration prediction elements from the target multi-source heterogeneous data includes: The target multi-source heterogeneous data is processed using multiple capabilities of the large language model to extract target triple data; A domain knowledge graph is constructed based on the target triplet data, and the performance of the domain knowledge graph in different dimensions is quantified to generate a quantitative predictive element set.

[0010] In conjunction with the first aspect, this application provides a third possible implementation of the first aspect, wherein the multiple approaches include at least knowledge mining; Mining mineral exploration prediction elements from the target multi-source heterogeneous data includes: Based on various mineral exploration-related factors in the area to be predicted, the mineral exploration prediction factors corresponding to different types of mineral deposits in the area to be predicted are analyzed. The mineral exploration prediction elements corresponding to the different mineral deposit types are organized into a target form to obtain the knowledge prediction element set.

[0011] In conjunction with the first aspect, this application provides a fourth possible implementation of the first aspect, wherein the multiple approaches include at least spatial mining; Mining mineral exploration prediction elements from the target multi-source heterogeneous data includes: Spatial geoscience data is extracted from the target multi-source heterogeneous data, and various types of feature layers are constructed based on the spatial geoscience data; Calculate the target parameter data of the various types of feature layers, and filter out key spatial variables based on the target parameter data to form a set of spatial prediction features.

[0012] In conjunction with the first aspect, this application provides a fifth possible implementation of the first aspect, wherein the fusion of the element sets corresponding to the multiple pathways to obtain the full prediction element system includes: Extract the target core elements and unique elements from the element sets corresponding to model mining, knowledge mining and spatial mining respectively; The unique elements are associated to construct a complete predictive element system with the target core elements after association.

[0013] In conjunction with the first aspect, embodiments of this application provide a sixth possible implementation of the first aspect, wherein, associated with the unique elements, it includes: The unique elements are associated with the nodes and relationships in the constructed domain knowledge graph through the domain ontology in the domain knowledge graph; The association results generated by the association are collected, and a computational knowledge framework containing geological logical relationships is built based on the association results in order to construct an intelligent mineral exploration prediction model with a complete prediction element system.

[0014] In conjunction with the first aspect, this application provides a seventh possible implementation of the first aspect, wherein the intelligent mineral exploration prediction model is continuously optimized based on the predicted results and actual exploration results, including: The geological rationality of the prediction results is assessed, and the prediction results are fine-tuned based on the assessment results to obtain the target prediction result; The target prediction result and the newly added exploration result trigger the target mechanism of the intelligent mineral exploration prediction model for continuous optimization.

[0015] Secondly, embodiments of this application provide a multi-pathway fusion intelligent mineral exploration prediction model construction device, the device comprising: The acquisition module is used to acquire multi-source heterogeneous data related to intelligent mineral exploration in the area to be predicted, and to preprocess the multi-source heterogeneous data according to the data type of the multi-source heterogeneous data to obtain target multi-source heterogeneous data. The configuration module is used to invoke pre-configured multiple approaches in parallel to mine mineral exploration prediction elements in the target multi-source heterogeneous data in parallel, and obtain the element sets corresponding to each approach; the multiple approaches include at least model mining, knowledge mining and spatial mining. A construction module is used to integrate the element sets corresponding to multiple approaches to obtain a full prediction element system, and to construct an intelligent mineral exploration prediction model based on the full prediction element system. The prediction module is used to make predictions based on the type of mineral deposit in the area to be predicted according to the intelligent mineral exploration prediction model to obtain interpretable prediction results, and to continuously optimize the intelligent mineral exploration prediction model in conjunction with the prediction results and new exploration results.

[0016] Thirdly, embodiments of this application provide an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, the steps of the multi-path fusion intelligent mineral exploration prediction model construction method described in any one of the claims are executed.

[0017] This application provides a method for constructing a multi-pathway fusion-based intelligent mineral exploration prediction model. The method first acquires multi-source heterogeneous data related to intelligent mineral exploration in the area to be predicted, and preprocesses the data according to its data type to obtain target multi-source heterogeneous data. Second, it calls pre-configured multiple pathways in parallel to mine mineral exploration prediction elements from the target multi-source heterogeneous data, obtaining element sets corresponding to each pathway. Then, it fuses these element sets to obtain a complete prediction element system, and constructs an intelligent mineral exploration prediction model based on this system. Finally, it uses the intelligent mineral exploration prediction model to predict the type of mineral deposit in the area to be predicted, obtaining interpretable prediction results. The method then continuously optimizes the intelligent mineral exploration prediction model based on the prediction results and newly added exploration results. Based on the above methods, this application establishes parallel multi-pathways and innovatively designs a knowledge computing and fusion center to achieve the standardization, computation, and deep semantic fusion of three types of heterogeneous information: expert experience, implicit knowledge from literature, and spatial data patterns. Ultimately, it constructs an interpretable, evolvable, and human-machine collaborative intelligent mineral exploration prediction model. It also realizes that with "knowledge computing and data-driven" as its core features, it can identify mineralization elements and mineralization control relationships from multi-source heterogeneous data, extract mineralization patterns from literature knowledge, and establish a semantic network. On this basis, it can further carry out predictive reasoning through machine learning or deep learning under knowledge constraints, and realize the intelligent discovery, interpretability, and continuous optimization of predictive elements. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A schematic diagram of the process of a mineral exploration model in the prior art is shown; Figure 2 The illustration shows a flowchart of a method for constructing a multi-pathway fusion intelligent mineral exploration prediction model according to an embodiment of this application; Figure 3 This invention provides a schematic diagram of a columnar structure during spatial excavation according to an embodiment of the present application. Figure 4 A schematic diagram of the fusion of the full prediction element system provided in an embodiment of this application is shown; Figure 5 This paper shows a structural block diagram of a multi-path fusion intelligent mineral exploration prediction model construction device provided in an embodiment of this application; Figure 6 A structural block diagram of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0021] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0023] Existing technologies have several significant limitations that prevent mineral exploration prediction models from meeting the requirements of intelligence, quantification, and interpretability. Therefore, developing more intelligent, quantitative, and interpretable mineral exploration prediction models has become an inevitable trend and core requirement for the development of modern mineral exploration technology.

[0024] Based on this, the embodiments of this application provide a method, apparatus and equipment for constructing a multi-pathway fusion intelligent mineral exploration prediction model, which will be described below through embodiments.

[0025] Example 1 To facilitate understanding of this embodiment, a method for constructing a multi-pathway fusion intelligent mineral exploration prediction model disclosed in this application will first be described in detail. For example... Figure 2 The diagram illustrates a method for constructing a multi-pathway fusion-based intelligent mineral exploration prediction model. This application provides a method for constructing a multi-pathway fusion-based intelligent mineral exploration prediction model, the method comprising: S101. Acquire multi-source heterogeneous data related to intelligent mineral exploration in the area to be predicted, and preprocess the multi-source heterogeneous data according to the data type of the multi-source heterogeneous data to obtain target multi-source heterogeneous data. S102. Parallel invocation of pre-configured multiple approaches to mine mineral exploration prediction elements in the target multi-source heterogeneous data in parallel, obtaining element sets corresponding to each of the multiple approaches; the multiple approaches include at least model mining, knowledge mining and spatial mining. S103. A comprehensive prediction element system is obtained by integrating the element sets corresponding to multiple approaches, and an intelligent mineral exploration prediction model is constructed based on the comprehensive prediction element system. S104. Based on the intelligent mineral exploration prediction model, predict the mineral deposit type of the area to be predicted to obtain interpretable prediction results, and continuously optimize the intelligent mineral exploration prediction model in conjunction with the prediction results and the newly added exploration results.

[0026] In step S101, this application acquires multi-source heterogeneous data related to intelligent mineral exploration in the area to be predicted. This multi-source heterogeneous data includes unstructured text, geological exploration data, and typical mineral deposits. Preprocessing is performed according to the data types of the multi-source heterogeneous data to obtain target multi-source heterogeneous data. Correspondingly, the target multi-source heterogeneous data includes geological literature and report text data, spatial geoscience data, and mineral deposit (point) data. Specifically, for geological literature and report text data, unstructured texts such as academic papers, exploration reports, and dissertations related to the area to be predicted are collected and converted into a structured text database through optical character recognition (OCR) and manual proofreading. The spatial geoscience data is the collected and standardized geological exploration data of the area to be predicted, including: ① Geological data: stratigraphy, lithology, structure, and distribution of igneous rocks; ② Geophysical data: aeromagnetic and gravity data and their anomaly information; ③ Geochemical data: geochemical measurement data such as soil sediments and stream sediments and their elemental anomaly information; ④ Remote sensing data: alteration anomaly information, etc.; the mineral deposit (point) data is a collection of attribute information such as geographical location, mineral deposit type, scale, and grade of typical mineral deposits in the study area, which can be used as sample labels for training and verification of the intelligent mineral exploration prediction model to improve training accuracy.

[0027] In step S102, this application pre-sets multiple approaches for mining mineral exploration prediction elements based on the data type of the target multi-source heterogeneous data. These multiple approaches specifically include model mining, knowledge mining, and spatial mining. Therefore, after obtaining the target multi-source heterogeneous data, the pre-configured multiple approaches are called in parallel to mine the mineral exploration prediction elements in the target multi-source heterogeneous data, resulting in element sets corresponding to each approach. That is, the number of mineral exploration prediction elements mined by each approach is multiple, thereby realizing the mining of mineral exploration prediction elements from different dimensions, breaking the limitations of a single mining method, ensuring the comprehensiveness of the mineral exploration prediction elements in the mined element sets, initially constructing a prediction knowledge system, providing a solid foundation for the subsequent construction of intelligent and quantitative mineral exploration prediction models, and specifically improving the problems of strong subjectivity, fragmented knowledge, and difficulty in comprehensively processing multi-source data in existing expert-driven models, thereby enhancing the comprehensiveness and scientific nature of mineral exploration prediction and meeting the core requirements of modern mineral exploration for prediction models.

[0028] It should be noted that the areas to be predicted and the types of mineral deposits are not limited here, and can be adjusted adaptively according to actual needs. For example, this embodiment is for the construction of a gold prospecting prediction model for the Qinling metallogenic belt, which is only used as an example and is not intended to be a limitation.

[0029] In a specific implementation of step S102, one embodiment is as follows: the pre-configured multiple paths are implemented based on the following steps: S1021. Pre-configure the multiple paths so that each of the multiple paths has a corresponding processing method; different paths correspond to different processing methods; S1022. The processing method is executed in parallel to mine mineral exploration prediction elements in the target multi-source heterogeneous data.

[0030] In steps S1021-S1022, this application pre-sets three approaches—model mining, knowledge mining, and spatial mining—based on the geological literature and report text data, spatial geoscience data, and mineral deposit (point) data included in the target multi-source heterogeneous data. It also sets specific processing methods for each approach, i.e., mining logic, processes, and methods adapted to the corresponding data types. This ensures that each approach has a clear and implementable operational plan, rather than random mining. The core logic and operational focus of the three approaches differ, and they also adapt to different data dimensions. For example, knowledge mining focuses on combining geological expert experience and mineralization theories to process qualitative geological knowledge; model mining focuses on processing quantitative data through algorithmic models; and spatial mining focuses on analyzing the spatial distribution and correlation characteristics of the data. The three processing methods complement each other, covering different types of data mining needs. This application does not use the three approaches sequentially, but instead initiates the processing methods corresponding to the three approaches simultaneously, and mines mineral exploration prediction elements from the target multi-source heterogeneous data. This not only efficiently processes massive heterogeneous data, but also mines mineral exploration prediction elements from different dimensions, ultimately forming multiple sets of elements, which provides support for building a prediction knowledge system and solving the limitations of existing technologies such as strong subjectivity and fragmentation.

[0031] In the specific implementation of step S1022, one embodiment is as follows: the multiple approaches include at least model mining; Mining mineral exploration prediction elements from the target multi-source heterogeneous data includes: A1. Apply the multiple capabilities of the large language model to process the target multi-source heterogeneous data in order to extract the target triple data; A2. Construct a domain knowledge graph based on the target triplet data, and quantify the performance of the domain knowledge graph in different dimensions to generate a quantitative predictive element set.

[0032] In steps A1-A2, the model set in the model mining approach of this application is a Large Language Model (LLM). The model mining approach constructs a prediction model based on the large model. This approach uses the text reading, induction, and reasoning capabilities of the Large Language Model (LLM) to replace expert reading of literature, so as to quickly extract knowledge graphs and realize the intelligent construction of gold mine prospecting prediction models. This solves the problem that with the advent of the information age, the number of documents that experts need to deal with is rapidly increasing, and it is difficult to quickly process a large number of papers and knowledge by relying solely on manual reading. First, this application designs multiple sets of dedicated prompts to control the large language model to automatically extract "deposit-attribute-relation" triplet data from the preprocessed text database established in step S101. It can also fine-tune traditional named entity recognition (NER) and relation extraction (RE) models based on pre-trained models. The prompt types include entity extraction prompts and relation extraction prompts for concepts, instances, phenomena, and numerical values. The extraction process calls the open-source Qwen2-72b-instruct model, which has high reusability. It can also be replaced by other general or domain-specific large language models with similar text understanding and generation capabilities, such as the GPT series, GLM series, and Claude series. Further... One step involves using a specialized model fine-tuned for geological text to improve the accuracy of extracting technical terms and complex expressions. Modifying some prompts allows for application to other mineral maps. This application also utilizes Retrieval Enhancement Generation (RAG) and intelligent agent mechanisms to perform semantic-level alignment and disambiguation on the extracted triplet data. Automated alignment can also be achieved using a deep semantic matching model based on a pre-trained Siamese network; alternatively, a more traditional method based on domain synonym lists and rule matching can be used as a lightweight alternative in scenarios with high terminology standardization. The processed target triplet data is stored in a graph database (such as Neo4j) to construct a structured domain knowledge graph. Furthermore, this application performs graph algorithms such as common node analysis and second-level weighted Jaccard similarity calculation on the domain knowledge graph to quantify and statistically analyze the commonalities and differences among different mineral deposit types in terms of mineral assemblages, alteration characteristics, and elemental anomalies. For example, by statistically analyzing the frequency of a certain characteristic node, such as "silicification," "fracture-controlled ore," or "intermediate-acidic magmatic activity," which is connected to all deposits under a certain type of mineral deposit (e.g., metamorphic volcanic-sedimentary rock type), the prevalence and importance of this node in that type of mineral deposit can be quantitatively assessed. Then, hierarchical clustering is used to calculate the similarity between different mineral deposits, thereby visualizing the similarity of multi-source information of mineral deposits and revealing the diversity, spatial differentiation, and genetic complexity of mineralization. Finally, the most significant common features of mineral deposits are selected to automatically generate a set of quantitative prediction elements corresponding to the model mining path, as shown in Table 1.

[0033] Table 1. Set of Quantitative Prediction Elements

[0034] In the specific implementation of step S1022, another embodiment exists in which the multiple approaches at least include knowledge mining; Mining mineral exploration prediction elements from the target multi-source heterogeneous data includes: B1. Based on various mineral exploration-related factors in the area to be predicted, analyze the mineral exploration prediction factors corresponding to different types of mineral deposits in the area to be predicted. B2. Organize the mineral exploration prediction elements corresponding to the different mineral deposit types into target form to obtain the knowledge prediction element set.

[0035] In steps B1-B2, the knowledge mining approach described in this application constructs a prediction model based on manual reading. This application utilizes geological experts to systematically study various mineral exploration-related factors corresponding to the literature and data of the area to be predicted. These factors include the classification principles of major gold deposits in China, the latest findings on gold exploration and research in the Qinling Mountains, and gold metallogenic theories. Specifically, based on the classification principles of major gold deposits in China and combined with the latest findings on gold exploration and research in the Qinling Mountains and gold metallogenic theories, different deposit types in the Qinling metallogenic belt are manually summarized and categorized, such as metamorphic volcanic-sedimentary rock type (greenschist type). For ease of use, the prospecting prediction elements for clastic rock type (Carlin-like), carbonate rock type (Carlin-like), and shallow intermediate-acidic intrusive rock type are organized into target form. That is, by describing the host rocks, mineralization type, ore-controlling elements, and metallogenic characteristics of each type of deposit in tabular form, the prospecting prediction elements are matched with the intelligent prospecting prediction model to obtain the set of knowledge prediction elements directly extracted by experts through knowledge mining, as shown in Table 2. This provides standardized and systematic knowledge support for the subsequent construction of intelligent prospecting prediction models.

[0036] Table 2. Collection of Knowledge Prediction Elements

[0037] In the specific implementation of step S1022, there is another embodiment in which the multiple approaches include at least spatial mining; Mining mineral exploration prediction elements from the target multi-source heterogeneous data includes: C1. Extract spatial geoscience data from the target multi-source heterogeneous data, and construct various types of feature layers based on the spatial geoscience data; C2. Calculate the target parameter data of the various types of feature layers, and filter out key spatial variables based on the target parameter data to form a set of spatial prediction elements.

[0038] In steps C1-C2, the spatial mining approach described in this application constructs a prediction model based on spatial data mining. This application first identifies spatial geoscience data within the multi-source heterogeneous data, extracts this spatial geoscience data, and performs in-depth analysis, including: statistically analyzing the spatial relationships between mineral deposits and strata, lithology, faults, and rock masses of different ages; analyzing the spatial distribution and correlation of geochemical elements; and extracting geophysical anomaly features, thereby constructing various types of feature layers, including stratigraphic age layers, igneous rock distribution layers, tectonic density layers, geochemical anomalies, and geophysical anomalies. All feature layers are used as initial features, such as... Figure 3 The input is fed into machine learning models with feature importance assessment capabilities, such as random forests or gradient boosting decision trees (e.g., XGBoost, LightGBM), support vector machines (with specific feature selection methods), or convolutional neural networks (CNN) and long short-term memory networks (LSTM). Using known ore deposit points as positive samples, the target parameter data of each feature layer, such as importance scores and AUC values, are calculated. Based on the target parameter data, key spatial variables are selected. These key spatial variables are spatial variables that contribute highly to mineralization prediction. A set of spatial prediction elements is formed using these key spatial variables, as shown in Table 3.

[0039] Table 3 Spatial Prediction Element Set Table

[0040] In step S103, when obtaining the element set corresponding to each path, this application fuses the element sets corresponding to multiple paths to obtain a complete prediction element system, such as... Figure 4 As shown, this specifically includes commonality extraction and cross-validation, difference complementarity and semantic fusion. It integrates quantitative elements from model mining, qualitative elements from knowledge mining, and spatial correlation elements from spatial mining into a comprehensive, unified, and non-redundant full-prediction element system. Based on this full-prediction element system, an intelligent mineral exploration prediction model is constructed. The full-prediction element system is the core foundation for building this intelligent mineral exploration prediction model. This system overcomes the limitations of existing technologies such as "fragmented knowledge" and "single elements," covering mineral exploration prediction elements of different dimensions and types, and possessing computable and spatially quantifiable characteristics. The intelligent mineral exploration prediction model built based on this system can break free from excessive reliance on individual expert experience, achieving quantitative and intelligent evaluation of mineral potential areas. This aligns with the core requirement of "developing intelligent, quantitative, and interpretable mineral exploration prediction models," while also compensating for many limitations of existing technologies.

[0041] This application employs three collaborative approaches to deeply mine target multi-source heterogeneous data: knowledge mining refines high-value experiences; model mining processes massive amounts of unstructured literature, automatically and on a large scale extracting and statistically analyzing implicit patterns; and spatial mining processes massive amounts of spatial grid data, uncovering nonlinear, high-dimensional spatial correlation patterns. The sets of elements mined by these three approaches are then fused, integrating information mined from different dimensions and data types. For example, spatial data might show a weak relationship between an anomaly of a certain element and mineralization (which traditional methods might ignore), but a knowledge graph reveals the crucial role of that element in a specific ore-forming fluid. Combining the two may uncover new, genetically significant weak anomaly prospecting indicators. This deep fusion significantly expands the sources and depth of effective information, enabling intelligent mineral exploration prediction models not only to fit known patterns but also to discover implicit knowledge and new patterns that are difficult for the human brain to directly summarize from complex data.

[0042] In a specific implementation of step S103, one embodiment is as follows: the fusion of the element sets corresponding to the multiple pathways to obtain the full prediction element system includes: S1031. Extract the target core elements and unique elements from the element sets corresponding to model mining, knowledge mining and spatial mining respectively. S1032. Associate the unique elements to construct a complete prediction element system with the target core elements after association.

[0043] In steps S1031-S1032, this application compares and analyzes the predicted element sets from three approaches, extracting the high-confidence target core elements commonly identified by the predicted element sets from the three approaches. For example, the Qinling gold mine exhibits "silicification alteration," "Au-As-Sb anomaly combination," and "fracture-controlled ore," meaning that the target core elements are common features in the predicted element sets from the three approaches, as shown in Table 4. Only when an element (such as "silicification") is supported by knowledge mining, model mining, and spatial mining is it confirmed as a high-confidence target core element. This multi-source mutual verification mechanism greatly reduces the subjective risk caused by the cognitive bias of a single information source or a single individual, achieving mutual verification. It also extracts the unique elements of each of the predicted element sets from the three approaches, as shown in Table 5, such as the detailed mineral coherence obtained based on the manual reading approach. The system uses a domain ontology to semantically link and fuse differentiated elements, including the generation sequence, implicit element coupling relationships identified through large language model approaches, and geophysical gradient zone information based on spatial data mining. For example, it links "potassium-rich feldspar alteration" (expert knowledge) with "high K anomaly zone" (spatial data mining), thereby fusing them to obtain a complete predictive element system. Based on this complete predictive element system, a structured intelligent mineral exploration prediction model is constructed for each target deposit type. This model is no longer a simple weighted table, but a computational knowledge framework with multiple modules including "geological features (ore-bearing host rocks, structures, minerals, alteration)," "geophysical features," and "geochemical features," and elements with semantic relationships between them. Each element under each module has a source identifier (which approach it comes from) and quantified contribution information.

[0044] Table 4. Common Characteristics of Core Target Elements

[0045] Table 5 Unique Elements Table

[0046] An alternative to the fusion of contrastive analysis and semantic ontology could be a more formalized uncertainty quantification and Bayesian fusion method. Each element extracted through a different path is assigned a credibility or probability score, and during fusion, a weighted synthesis is performed based on the reliability and consistency of its source, thus handling conflicting information more precisely. Figure 4 The fusion sequence shown involves three parallel outputs for fusion. An alternative is to use sequential or iterative fusion. For example, first, a large model can be used to quickly organize literature to form a preliminary knowledge framework, guiding experts to make more targeted experience summaries (human-computer collaborative reading), and then using this as a guide for targeted mining and analysis of spatial data. Another alternative is to construct a hierarchical structure that fuses two data points at a time before fusioning the whole data.

[0047] The alternatives described in this application indicate that the scope of protection of this invention is not limited to the specific tools, algorithms, or data sources mentioned in the embodiments. Any variations and combinations of the core methodological framework of "extracting prediction information from three heterogeneous sources—expert experience, unstructured text, and spatial data—and generating a structured, interpretable, comprehensive prediction model through a computational fusion process" should be considered within the scope of protection of this invention.

[0048] In a specific implementation of step S1032, one embodiment includes associating the unique elements, including: S10321. Associate the unique elements with the nodes and relationships in the constructed domain knowledge graph through the domain ontology in the domain knowledge graph; S10322. Collect the association results generated by the association, and build a computational knowledge framework containing geological logical relationships based on the association results, so as to construct an intelligent mineral exploration prediction model with a complete prediction element system.

[0049] In steps S10321-S10322, this application associates the unique elements with the nodes and relationships in the constructed domain knowledge graph through the domain ontology in the domain knowledge graph. The domain ontology is the core bridge connecting the unique elements and the domain knowledge graph; essentially, it is a set of standardized geological domain semantic rules to clarify the definitions and association logic of various elements and nodes. Through these rules, the unique elements in the entire prediction element system are precisely associated with the nodes and relationships between nodes in the domain knowledge graph. For example, the "alteration marker of porphyry copper deposit" (a unique element) is bound to the "porphyry copper deposit" node and the "alteration marker" node in the knowledge graph, as well as their corresponding relationship. This association process generates a series of association results, such as the matching relationship between unique elements and knowledge graph nodes, and the indirect associations formed by different unique elements through knowledge graph nodes, such as element A associating with node B, node B associating with element C, thus forming an indirect geological association between A and C. These association results are collected. This is the foundation for building a computational knowledge framework. Based on the collected correlation results, a computational knowledge framework containing geological logical relationships is constructed. This computational knowledge framework integrates unique elements, nodes and relationships of the knowledge graph according to real geological logic into a quantifiable, reasonable, and computable whole. For example, if the mineralization correlation between different elements can be calculated through the framework, the constructed intelligent mineral exploration prediction model is based on the construction of a complete prediction element system, that is, a feature-integrated intelligent mineral exploration prediction model, as shown in Table 6. The computational knowledge framework is also the core architecture of the intelligent mineral exploration prediction model and an important component of the model. The intelligent mineral exploration prediction model constructed in this application can break away from the dependence on expert subjective experience and achieve intelligent and quantitative mineral exploration prediction based on geological logic.

[0050] Table 6 Intelligent Mineral Exploration Prediction Model

[0051] The domain knowledge graph constructed in this application based on model mining is essentially a semantic network. It explicitly expresses the diverse relationships between entities such as "strata," "structures," "igneous rocks," "alteration," "minerals," and "elements" in a structured manner, such as "originating from," "controlled by," and "associated with." Further, this step utilizes domain ontology to associate the mineral exploration prediction elements discovered through knowledge mining and spatial mining with the nodes and relationships in the knowledge graph. This makes the final intelligent mineral exploration prediction model not just a list of elements, but a computational knowledge framework containing geological logical relationships. For example, the intelligent mineral exploration prediction model can explain why a "high-K anomaly zone" is a mineral exploration indicator because it is associated with ore-forming hydrothermal activity through the "potassium feldspar alteration" node. This deep association endows the model with systematic explanatory power, enabling it to move from isolated phenomenon descriptions to an understanding of the mineralization process and supporting more complex associative reasoning.

[0052] This application constructs an organic collaborative workflow: experts play a core role in knowledge mining and the verification and intervention of intelligent mineral exploration prediction models, ensuring the correctness of the geological causal logic and the reliability of the results. Meanwhile, the arduous information processing in between, such as literature review, data mining, and complex calculations involving map construction and feature selection, is efficiently completed by large models and machine learning algorithms. Therefore, expert knowledge, through the intermediary of ontology and knowledge graphs, is transformed into constraints that computers can understand and process, thus guiding and interpreting the data-driven process rather than being replaced by it. This model, where "experts set the framework and rules, and machines execute calculations and explorations," avoids the "algorithm black box" and liberates expert productivity. The resulting model is a hybrid of human and machine intelligence, whose prediction results possess both the objective breadth of data-driven approaches and the logical depth of expert knowledge-driven approaches, significantly enhancing interpretability and practicality.

[0053] In step S104, after constructing the intelligent mineral exploration prediction model, this application applies the intelligent mineral exploration prediction model to the area to be predicted, makes predictions based on the mineral deposit type of the area to be predicted, generates a mineral exploration favorable zoning map in the GIS platform, delineates the mineral exploration target area, and thus obtains interpretable prediction results. In order to achieve synergy between the intelligent mineral exploration prediction model construction technology and the actual mineral exploration prediction application, this application proposes a closed-loop mechanism of model application—expert verification—model feedback—knowledge evolution, so as to continuously optimize the intelligent mineral exploration prediction model based on the prediction results and the newly collected exploration results corresponding to the prediction results, thereby ensuring the accuracy of the intelligent mineral exploration prediction model. The newly added exploration results include newly discovered mineral points and borehole data.

[0054] In summary, this invention establishes three parallel information extraction pathways and innovatively designs a knowledge computing and fusion center to achieve the standardization, computation, and deep semantic fusion of three types of heterogeneous information: expert experience, implicit knowledge from documents, and spatial data patterns. Ultimately, it constructs an interpretable, evolvable, and human-machine collaborative intelligent mineral exploration prediction model.

[0055] In the specific implementation of step S104, one embodiment is as follows: Based on the prediction results and actual exploration results, the intelligent mineral exploration prediction model is continuously optimized, including: S1041. The geological rationality of the prediction results is judged, and the prediction results are fine-tuned based on the judgment results to obtain the target prediction results; S1042. Based on the target prediction results and the newly added exploration results, the target mechanism of the intelligent mineral exploration prediction model is triggered for continuous optimization.

[0056] In steps S1041-S1042, this application uses geological experts to assess the geological rationality of the prediction results and can fine-tune the target area boundaries or priorities based on their experience to obtain the target prediction results. Simultaneously, experts review and correct the key entities and relationships in the knowledge graph generated through the model's mining process. Then, new exploration results in the area to be predicted, such as newly discovered mineral deposits, borehole data, and the target prediction results containing expert corrections, are used as new input data and fed back to the data preprocessing stage and knowledge graph of the intelligent mineral exploration prediction model, triggering a target mechanism. This target can be an incremental learning mechanism or an event-based mechanism. The driven real-time or near-real-time update mechanism, such as when the number of new instances of a certain type of mineral deposit in the knowledge graph reaches a certain level, or when there is a major update to the spatial database, automatically triggers the incremental learning and optimization process of the relevant prediction model. This triggers the optimization of machine learning model parameters and the completion of the knowledge graph, thereby achieving the adaptive and continuous evolution of the mineral exploration prediction model. New exploration results and target prediction results containing expert corrections serve as feedback signals, which can not only be directly used to optimize the intelligent mineral exploration prediction model, such as updating training samples, but more importantly, can be fed back into the knowledge graph and domain ontology. This feedback enables incremental supplementation and correction of structured knowledge. For example, the discovery of a new mineral deposit and its characteristics are extracted as triples and added to the graph, which may strengthen or correct the weights of certain element associations; an expert's negation of a reasoning chain can trigger adjustments to the ontology rules. This transforms the entire intelligent mineral exploration prediction model from a static "knowledge snapshot" into a "living system" with learning capabilities, capable of co-evolving with the deepening of geological understanding and the accumulation of exploration data, fundamentally solving the problem of lagging traditional model updates.

[0057] One alternative to expert correction is to design a more structured expert feedback interface that allows experts to label samples that are misclassified by the machine learning model, question the ranking of feature importance and retrain, or directly edit the fusion rules (the logic within the fusion center). Example 2 This application also provides a device for constructing a multi-pathway integrated intelligent mineral exploration prediction model, such as... Figure 5 The diagram shows a block diagram of a multi-pathway fusion intelligent mineral exploration prediction model construction device. The functions implemented by this device correspond to the steps of executing a multi-pathway fusion intelligent mineral exploration prediction model construction method on a terminal device as described above. This device can be understood as a server component including a processor. The multi-pathway fusion intelligent mineral exploration prediction model construction device described in this application includes: The acquisition module 501 is used to acquire multi-source heterogeneous data related to intelligent mineral exploration in the area to be predicted, and to preprocess the multi-source heterogeneous data according to the data type of the multi-source heterogeneous data to obtain target multi-source heterogeneous data. Configuration module 502 is used to call pre-configured multiple approaches in parallel to mine mineral exploration prediction elements in the target multi-source heterogeneous data in parallel, and obtain the element sets corresponding to the multiple approaches respectively; the multiple approaches include at least model mining, knowledge mining and spatial mining. The construction module 503 is used to integrate the element sets corresponding to multiple paths to obtain a full prediction element system, and to construct an intelligent mineral exploration prediction model based on the full prediction element system. The prediction module 504 is used to make predictions based on the type of mineral deposit in the area to be predicted according to the intelligent mineral exploration prediction model to obtain interpretable prediction results, and to continuously optimize the intelligent mineral exploration prediction model in conjunction with the prediction results and new exploration results.

[0058] In one feasible implementation, the configuration module includes: The multiple pathways are pre-configured so that each pathway has a corresponding processing method; different pathways correspond to different processing methods. The processing method is executed in parallel to mine mineral exploration prediction elements in the target multi-source heterogeneous data.

[0059] In one feasible implementation, the configuration module further includes: The target multi-source heterogeneous data is processed using multiple capabilities of the large language model to extract target triple data; A domain knowledge graph is constructed based on the target triplet data, and the performance of the domain knowledge graph in different dimensions is quantified to generate a quantitative predictive element set.

[0060] In one feasible implementation, the configuration module also includes: Based on various mineral exploration-related factors in the area to be predicted, the mineral exploration prediction factors corresponding to different types of mineral deposits in the area to be predicted are analyzed. The mineral exploration prediction elements corresponding to the different mineral deposit types are organized into a target form to obtain the knowledge prediction element set.

[0061] In one feasible implementation, the configuration module further includes: Spatial geoscience data is extracted from the target multi-source heterogeneous data, and various types of feature layers are constructed based on the spatial geoscience data; Calculate the target parameter data of the various types of feature layers, and filter out key spatial variables based on the target parameter data to form a set of spatial prediction features.

[0062] In one feasible implementation, the building module includes: Extract the target core elements and unique elements from the element sets corresponding to model mining, knowledge mining and spatial mining respectively; The unique elements are associated to construct a complete predictive element system with the target core elements after association.

[0063] In one feasible implementation, the building module further includes: The unique elements are associated with the nodes and relationships in the constructed domain knowledge graph through the domain ontology in the domain knowledge graph; The association results generated by the association are collected, and a computational knowledge framework containing geological logical relationships is built based on the association results in order to construct an intelligent mineral exploration prediction model with a complete prediction element system.

[0064] In one feasible implementation, the prediction module includes: The geological rationality of the prediction results is assessed, and the prediction results are fine-tuned based on the assessment results to obtain the target prediction result; The target prediction result and the newly added exploration result trigger the target mechanism of the intelligent mineral exploration prediction model for continuous optimization.

[0065] Example 3 This application also provides an electronic device, such as Figure 6As shown, it includes: a processor 601, a memory 602, and a bus 603. The memory 602 stores machine-readable instructions that can be executed by the processor 601. When the electronic device is running, the processor 601 and the memory 602 communicate through the bus 603. When the machine-readable instructions are executed by the processor 601, the steps of any one of the methods for constructing a multi-path fusion intelligent mineral exploration prediction model are performed.

[0066] Example 4 This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of any one of the methods for constructing a multi-pathway fusion intelligent mineral exploration prediction model.

[0067] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some communication interfaces; the indirect coupling or communication connection of devices or modules can be electrical, mechanical, or other forms.

[0068] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0069] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0070] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a platform server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0071] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for constructing a multi-pathway fusion intelligent mineral exploration prediction model, characterized in that, The method includes: Acquire multi-source heterogeneous data related to intelligent mineral exploration in the area to be predicted, and preprocess the multi-source heterogeneous data according to the data type of the multi-source heterogeneous data to obtain target multi-source heterogeneous data; The system invokes pre-configured multiple approaches in parallel to mine mineral exploration prediction elements in the target multi-source heterogeneous data, obtaining element sets corresponding to each approach; the multiple approaches include at least model mining, knowledge mining, and spatial mining. A comprehensive prediction element system is obtained by integrating the element sets corresponding to multiple approaches, and an intelligent mineral exploration prediction model is constructed based on the comprehensive prediction element system. Based on the intelligent mineral exploration prediction model, interpretable prediction results are obtained according to the type of mineral deposit in the area to be predicted. The intelligent mineral exploration prediction model is then continuously optimized in conjunction with the prediction results and new exploration results.

2. The method according to claim 1, characterized in that, The pre-configured multiple pathways are implemented based on the following steps: The multiple pathways are pre-configured so that each pathway has a corresponding processing method; different pathways correspond to different processing methods. The processing method is executed in parallel to mine mineral exploration prediction elements in the target multi-source heterogeneous data.

3. The method according to claim 2, characterized in that, The multiple approaches include at least model mining; Mining mineral exploration prediction elements from the target multi-source heterogeneous data includes: The target multi-source heterogeneous data is processed using multiple capabilities of the large language model to extract target triple data; A domain knowledge graph is constructed based on the target triplet data, and the performance of the domain knowledge graph in different dimensions is quantified to generate a quantitative predictive element set.

4. The method according to claim 2, characterized in that, The multiple approaches include at least knowledge mining; Mining mineral exploration prediction elements from the target multi-source heterogeneous data includes: Based on various mineral exploration-related factors in the area to be predicted, the mineral exploration prediction factors corresponding to different types of mineral deposits in the area to be predicted are analyzed. The mineral exploration prediction elements corresponding to the different mineral deposit types are organized into a target form to obtain the knowledge prediction element set.

5. The method according to claim 2, characterized in that, The multiple approaches include at least spatial mining; Mining mineral exploration prediction elements from the target multi-source heterogeneous data includes: Spatial geoscience data is extracted from the target multi-source heterogeneous data, and various types of feature layers are constructed based on the spatial geoscience data; Calculate the target parameter data of the various types of feature layers, and filter out key spatial variables based on the target parameter data to form a set of spatial prediction features.

6. The method according to claim 1, characterized in that, The fusion of the element sets corresponding to the multiple pathways yields a complete predictive element system, including: Extract the target core elements and unique elements from the element sets corresponding to model mining, knowledge mining and spatial mining respectively; The unique elements are associated to construct a complete predictive element system with the target core elements after association.

7. The method according to claim 6, characterized in that, The associated unique elements include: The unique elements are associated with the nodes and relationships in the constructed domain knowledge graph through the domain ontology in the domain knowledge graph; The association results generated by the association are collected, and a computational knowledge framework containing geological logical relationships is built based on the association results in order to construct an intelligent mineral exploration prediction model with a complete prediction element system.

8. The method according to claim 1, characterized in that, The intelligent mineral exploration prediction model is continuously optimized based on the combined prediction results and actual exploration results, including: The geological rationality of the prediction results is assessed, and the prediction results are fine-tuned based on the assessment results to obtain the target prediction result; The target prediction result and the newly added exploration result trigger the target mechanism of the intelligent mineral exploration prediction model for continuous optimization.

9. A device for constructing a multi-pathway fusion intelligent mineral exploration prediction model, characterized in that, The device includes: The acquisition module is used to acquire multi-source heterogeneous data related to intelligent mineral exploration in the area to be predicted, and to preprocess the multi-source heterogeneous data according to the data type of the multi-source heterogeneous data to obtain target multi-source heterogeneous data. The configuration module is used to invoke pre-configured multiple approaches in parallel to mine mineral exploration prediction elements in the target multi-source heterogeneous data in parallel, and obtain the element sets corresponding to each approach; the multiple approaches include at least model mining, knowledge mining and spatial mining. A construction module is used to integrate the element sets corresponding to multiple approaches to obtain a full prediction element system, and to construct an intelligent mineral exploration prediction model based on the full prediction element system. The prediction module is used to make predictions based on the type of mineral deposit in the area to be predicted according to the intelligent mineral exploration prediction model to obtain interpretable prediction results, and to continuously optimize the intelligent mineral exploration prediction model in conjunction with the prediction results and new exploration results.

10. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, they perform the steps of a multi-path fusion intelligent mineral exploration prediction model construction method as described in any one of claims 1 to 8.

Images