CN112927767B - Multi-element geochemical anomaly identification method based on graph attention self-coding - Google Patents
Multi-element geochemical anomaly identification method based on graph attention self-coding Download PDFInfo
- Publication number
- CN112927767B CN112927767B CN202110196133.0A CN202110196133A CN112927767B CN 112927767 B CN112927767 B CN 112927767B CN 202110196133 A CN202110196133 A CN 202110196133A CN 112927767 B CN112927767 B CN 112927767B
- Authority
- CN
- China
- Prior art keywords
- geochemical
- graph
- multivariate
- self
- graph attention
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/90—Programming languages; Computing architectures; Database systems; Data warehousing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/29—Geographical information databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/70—Machine learning, data mining or chemometrics
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Crystallography & Structural Chemistry (AREA)
- General Health & Medical Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Evolutionary Biology (AREA)
- Medical Informatics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Remote Sensing (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a multivariate geochemical anomaly identification method based on graph attention self-coding, which comprises the following steps of: determining a composition threshold value K; constructing a geochemical element topological network; learning the multivariate geochemical characteristics; reconstructing a multi-element geochemical element background; and calculating multiple abnormal values. The invention introduces graph learning into chemical exploration anomaly detection, utilizes K neighbor to construct a geochemical element topological relation graph, and constructs and trains a graph attention self-encoder capable of simultaneously extracting element composition relation and space structure characteristics. And reconstructing a multi-element geochemical background based on the trained graph attention self-encoder, and calculating to obtain the abnormal value of each final sampling point. The invention expands the existing neural network model, so that the model can directly process sampling point data and can be applied to irregular areas, the learning performance of a geophysical prospecting background and the practicability of the model are greatly improved, and a practical and reliable geophysical prospecting abnormity identification method is provided for complex geological conditions.
Description
Technical Field
The invention relates to the field of multivariate geochemical exploration anomaly identification and the field of artificial intelligence application, in particular to a multivariate geochemical anomaly identification method based on graph attention self-coding.
Background
Identification of multiple geochemical anomalies is one of important contents for mineral resource exploration, and anomaly information of the multiple geochemical anomalies helps geologists judge potential mineral deposits. The geochemical exploration anomaly has strong spatial heterogeneity, and the spatial heterogeneity of the geochemical exploration elements must be considered. The traditional anomaly identification methods such as fractal/multiple analysis, a kriging method, space factor analysis and the like consider the correlation of space neighbor samples and are remarkable in the identification of the chemical exploration anomalies. In recent years, convolutional neural networks have been introduced into the research field of geochemical anomaly identification and have achieved higher accuracy by virtue of having good and automatic learning capability for complex spatial features. The input data of the model must be based on grid data after the interpolation of the sampling points; this causes secondary errors in the data and also limits the application of the model in natural environments (irregularities). In addition, the convolution window for extracting the spatial structure features can only be rectangular and has no rotation invariance, which is not consistent with the distribution condition of element features in a real environment and also limits the capability of the convolution layer for extracting the spatial features. Therefore, there is a need for improving and expanding the existing convolutional self-coding, and solving the above problems to improve its capability of diversified background learning and anomaly identification.
Disclosure of Invention
The invention provides a multivariate geochemical anomaly recognition method based on graph attention self-coding, aiming at the problems that the prior convolutional self-encoder can not directly use sampling point data and can not consider real sampling points and irregular mining areas, and comprising the following steps:
s1, calculating the spatial correlation of the earth chemical elements under different distance thresholds, drawing a distance threshold and a line graph of the correlation, and selecting the distance threshold at the obvious turning point of the line graph as K of the K neighbor algorithm;
s2, normalizing the concentration data of the multiple geochemical elements, and performing edge connection on the multiple geochemical exploration sampling points based on a K neighbor algorithm to complete a topological graph;
s3, constructing a graph attention self-encoder model, training by using the topological graph in S2, and selecting a model parameter with the minimum reconstruction error as a final graph attention self-encoder model;
s4, inputting the topological graph in the S2 into a final graph attention self-encoder model obtained by S3 training, and outputting to obtain a background value of the multiple geochemical elements;
and S5, calculating Euclidean distances between the multi-element geochemical element concentration data and the multi-element geochemical element background values in the S2 to be used as abnormal values, mapping the abnormal values to a geographic space, and generating a multi-element abnormal exploration map.
Compared with the prior art, the invention has the following beneficial effects:
(1) the optimal distance threshold value K is determined by utilizing the Moran' I index, and the sampling points are topologically structured by utilizing the K neighbor algorithm, so that the sampling point data has attribute information and topological structure information and can be used for deep learning of the image.
(2) The method introduces the graph deep learning into the field of the chemical exploration anomaly identification, constructs a graph attention self-coding neural network, and provides a chemical exploration anomaly identification model considering real sampling points and irregular mining areas; the network model directly takes real geochemical sampling point data as input data, secondary errors caused by interpolation are avoided, the model can be applied to irregular areas, and the practicability of the geochemical model is improved.
(3) The structure responsible for space extraction in the self-coding constructed by the method is a graph attention layer, and the attribute characteristics and the structural characteristics of sampling points in the multi-element geochemical topological graph can be extracted; compared with the common convolutional layer, the convolutional layer has rotational invariance, can better extract the spatial characteristics of different regions and different structures, and provides an efficient method for carrying out anomaly identification and mineral deposit judgment by utilizing diversified probe data under complex geological and mineral-forming environments.
Drawings
The invention will be further described with reference to the accompanying drawings and examples, in which:
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a plot of data for 5 chemical elements of a Minxife ore band in accordance with an embodiment of the present invention;
FIG. 3 is a schematic diagram of the proposed detection anomaly identification based on graph attention self-coding;
FIG. 4 is a final anomaly result graph in an implementation example;
FIG. 5 is a graph showing comparative evaluation of model performance according to the present invention.
Detailed Description
For a more clear understanding of the technical features, objects and effects of the present invention, embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
Referring to fig. 1, the method for recognizing multivariate geochemical anomalies based on graph attention self-coding of the present invention comprises the following steps:
s1, geological and mineralizing environment according to research areaSelecting geochemical element concentration data which is properly related to the detection of mineral product abnormity, and normalizing the geochemical element concentration data, wherein the final processing result is shown in a figure 2; in this example, the mineral product is detected as Fe, and the geochemical element which is related to the abnormality is Cu-Zn-Mn-Pb-Fe2O3；
S2, carrying out qualitative measurement on the spatial correlation and the aggregation of the geochemical elements by utilizing a Moran' I index, and selecting a threshold inflection point which can reserve enough spatial heterogeneity and contains enough background samples as an optimal K value of a K nearest neighbor algorithm;
s3, traversing all sampling points by using a K neighbor algorithm based on the determined K value, and performing edge connection establishment on neighbor sampling points within the distance K to form a multi-element geochemical topological graph; wherein, the sampling points are converted into nodes of a topological graph, and the concentration of multiple geochemical elements collected by the sampling points is used as a node attribute vector;
s4, constructing a graph attention self-coding model shown in the figure 3, inputting the obtained multivariate geochemical topological graph constructed in the S3 into the model for model training, observing reconstruction loss, stopping training when the error between input data and output data is not changed any more, and obtaining a completely trained graph attention self-coding model, wherein a loss function is specifically as follows:
n is the number of nodes (sampling points) in the graph, F is the characteristic number of the nodes,
is the kth original feature (element value) of the ith node (sample point),
for the kth reconstructed feature (element background value) of the ith node.
S5, as shown in FIG. 3, inputting the obtained multivariate geochemical topological graph constructed in the S3 into the model trained in the S4 again, and outputting to obtain a multivariate geochemical exploration background value of the region;
s6, calculating the Euclidean distance between the original data and the reconstructed background value of each sample as an abnormal value, wherein the formula is as follows:
f is the characteristic number of the node,
is the kth original feature (element value) of the ith node (sample point),
for the kth reconstructed feature (element background value) of the ith node.
And mapping the abnormal value to the geographical position of the sample to obtain a final regional mapping abnormal recognition map, as shown in fig. 4.
Evaluating an abnormal map of a known mining point area by adopting an ROC curve; when the area under the ROC curve to the whole graph area ratio (AUC) is greater than 0.5, it is demonstrated that the model is usable, with better model performance as the AUC approaches 1. The AUC value of the self-coding model constructed in this study was 0.797, which is shown in fig. 5, and is superior to other methods for identifying chemical anomalies.
The invention introduces graph learning into the abnormal identification of chemical exploration, and provides a graph attention self-coding model considering real sampling points and irregular mining areas; the network model directly takes the real geochemical sampling point data as input data, thereby avoiding secondary errors caused by interpolation, enabling the model to be applied in irregular areas and improving the practicability of the geochemical model. In addition, the structure responsible for space extraction in the self-encoding constructed by the method is an attention layer of the drawing, and the attribute characteristics and the structural characteristics of sampling points in the multi-element geochemical topological graph can be extracted; compared with the common convolutional layer, the convolutional layer has rotational invariance, can better extract the spatial characteristics of different regions and different structures, and provides an efficient method for carrying out anomaly identification and mineral deposit judgment by utilizing diversified probe data under complex geological and mineral-forming environments.
While the present invention has been described with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, which are illustrative and not restrictive, and it will be apparent to those skilled in the art that various changes and modifications can be made therein without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (6)
1. The method for identifying the multivariate geochemical anomaly based on the graph attention self-coding is characterized by comprising the following steps of:
s1, calculating the spatial correlation of the earth chemical elements under different distance thresholds, drawing a distance threshold and a line graph of the correlation, and selecting the distance threshold at the obvious turning point of the line graph as K of the K neighbor algorithm;
s2, normalizing the concentration data of the multiple geochemical elements, and performing edge connection on the multiple geochemical exploration sampling points based on a K neighbor algorithm to complete a topological graph;
s3, constructing a graph attention self-encoder model, training by using the topological graph in S2, and selecting a model parameter with the minimum reconstruction error as a final graph attention self-encoder model;
s4, inputting the topological graph in the S2 into a final graph attention self-encoder model obtained by S3 training, and outputting to obtain a background value of the multiple geochemical elements;
and S5, calculating Euclidean distances between the multi-element geochemical element concentration data and the multi-element geochemical element background values in the S2 to be used as abnormal values, mapping the abnormal values to a geographic space, and generating a multi-element abnormal exploration map.
2. The method for identifying multivariate geochemical abnormalities based on graph attention self-encoding according to claim 1, wherein the step S1 is specifically to quantitatively measure the spatial correlation and aggregative property of geochemical elements by using Moran' I index, and finally select the optimal K value.
3. The method for identifying multivariate geochemical anomaly based on graph attention self-encoding according to claim 1, wherein the sampling points are converted into nodes of a topological graph in step S2, and the concentrations of the multivariate geochemical elements collected by the sampling points are used as node attribute vectors.
4. The method for multivariate geochemical anomaly recognition based on graph attention self-coding according to claim 1, wherein in step S3, the encoder and the decoder of the graph attention self-coder model are both composed of two graph convolutional layers with attention mechanism.
5. The method for multivariate geochemical anomaly recognition based on graph attention self-coding according to claim 1, wherein in step S3, the final graph attention self-coder model reconstruction error LREThe calculation formula is as follows:
wherein N is the number of sampling points, F is the characteristic number of the sampling points,
for the kth original feature of the ith sample point,
the k element background value of the ith sample point.
6. The method for multivariate geochemical anomaly recognition based on graph attention self-coding according to claim 1, wherein the abnormal value calculation formula in step S5 is as follows:
wherein F is the characteristic number of the sampling points,
for the kth original feature of the ith sample point,
is the k element background value of the ith node.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110196133.0A CN112927767B (en) | 2021-02-22 | 2021-02-22 | Multi-element geochemical anomaly identification method based on graph attention self-coding |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110196133.0A CN112927767B (en) | 2021-02-22 | 2021-02-22 | Multi-element geochemical anomaly identification method based on graph attention self-coding |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112927767A CN112927767A (en) | 2021-06-08 |
| CN112927767B true CN112927767B (en) | 2022-05-13 |
Family
ID=76170011
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110196133.0A Active CN112927767B (en) | 2021-02-22 | 2021-02-22 | Multi-element geochemical anomaly identification method based on graph attention self-coding |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112927767B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115458071B (en) * | 2022-08-08 | 2024-02-02 | 北京信息科技大学 | Method, device and equipment for predicting heavy metal content in soil |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015022806A1 (en) * | 2013-08-14 | 2015-02-19 | 独立行政法人石油天然ガス・金属鉱物資源機構 | Crust data analysis method, crust data analysis program, and crust data analysis device |
| CN107885966A (en) * | 2017-10-23 | 2018-04-06 | 中国地质大学（武汉） | The SVM abnormal chemical single element sorting techniques containing constraint |
| CN108710777A (en) * | 2018-05-21 | 2018-10-26 | 中国地质大学（武汉） | Abnormality recognition method is visited in the diversification that own coding neural network is accumulated based on multireel |
| CN111639067A (en) * | 2020-05-21 | 2020-09-08 | 中国地质大学（武汉） | Multi-feature fusion convolution self-coding multivariate geochemical anomaly identification method |
-
2021
- 2021-02-22 CN CN202110196133.0A patent/CN112927767B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015022806A1 (en) * | 2013-08-14 | 2015-02-19 | 独立行政法人石油天然ガス・金属鉱物資源機構 | Crust data analysis method, crust data analysis program, and crust data analysis device |
| CN107885966A (en) * | 2017-10-23 | 2018-04-06 | 中国地质大学（武汉） | The SVM abnormal chemical single element sorting techniques containing constraint |
| CN108710777A (en) * | 2018-05-21 | 2018-10-26 | 中国地质大学（武汉） | Abnormality recognition method is visited in the diversification that own coding neural network is accumulated based on multireel |
| CN111639067A (en) * | 2020-05-21 | 2020-09-08 | 中国地质大学（武汉） | Multi-feature fusion convolution self-coding multivariate geochemical anomaly identification method |
Non-Patent Citations (1)
| Title |
|---|
| 基于多维分形模型与指示克里格方法的地球化学异常识别研究;李晓晖等;《地理与地理信息科学》;20111115(第06期);第23-27页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112927767A (en) | 2021-06-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhu et al. | Intelligent logging lithological interpretation with convolution neural networks | |
| US9279905B2 (en) | Method for extracting a thumbnail image from a training image so as to constrain the multipoint geostatistical modeling of the subsoil | |
| CN111639067B (en) | Multi-feature fusion convolution self-coding multi-element geochemical anomaly identification method | |
| CN108710777B (en) | Diversified anomaly detection identification method based on multi-convolution self-coding neural network | |
| Jung et al. | Geological model sampling using PCA-assisted support vector machine for reliable channel reservoir characterization | |
| CN114994759A (en) | Intelligent carbon seal storage box identification method and system based on GAN network | |
| CN109763814B (en) | Stratum matching visual analysis method based on multi-dimensional logging data | |
| Silversides et al. | Automated recognition of stratigraphic marker shales from geophysical logs in iron ore deposits | |
| US20230161061A1 (en) | Structured representations of subsurface features for hydrocarbon system and geological reasoning | |
| CN112927767B (en) | Multi-element geochemical anomaly identification method based on graph attention self-coding | |
| Nikitin et al. | Hybrid and automated machine learning approaches for oil fields development: The case study of Volve field, North Sea | |
| Shan et al. | Identification of flow units using the joint of WT and LSSVM based on FZI in a heterogeneous carbonate reservoir | |
| GB2612694A (en) | System and method for probability-based determination of stratigraphic anomalies in a subsurface | |
| Friedel et al. | Hybrid modeling of spatial continuity for application to numerical inverse problems | |
| Li et al. | CNN-Based Network Application for Petrophysical Parameter Inversion: Sensitivity Analysis of Input–Output Parameters and Network Architecture | |
| Fakhari et al. | Fisher Discriminant Analysis (FDA), a supervised feature reduction method in seismic object detection | |
| CN113688901B (en) | Reservoir discontinuous boundary line identification method based on expansion convolutional neural network | |
| Roden et al. | Seismic interpretation with machine learning | |
| CN115964667A (en) | River-lake lithofacies well logging identification method based on deep learning and resampling | |
| CN115808710A (en) | Method for extracting pre-stack seismic data features of tight reservoir | |
| Liu et al. | Machine learning assisted recovery of subsurface energy: a review | |
| Tong et al. | Coal structure identification based on geophysical logging data: Insights from Wavelet Transform (WT) and Particle Swarm Optimization Support Vector Machine (PSO-SVM) algorithms | |
| Ren et al. | Reconstruction of Missing Well-Logs Using Facies-Informed Discrete Wavelet Transform and Time Series Regression | |
| Osogba | Machine Learning for Subsurface Data Analysis: Applications in Outlier Detection, Signal Synthesis and Core & Completion Data Analysis | |
| Mojeddifar et al. | Gas-bearing reservoir characterization using an adjusted Parzen probabilistic network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |