A geochemical anomaly lower limit determination method and system
Patent Information
- Application Number
- CN202610725153.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-05-25
- Publication Date
- 2026-10-16
AI Technical Summary
在地球化学找矿中，与成矿密切相关元素的变异系数通常较大，其含量变化对识别矿化异常具有更高的指示价值，而等权重处理会削弱这些关键元素的贡献，降低异常识别的灵敏度
[0023] The method for determining the lower limit of geochemical anomalies provided in this application has the following advantages: It eliminates closure effects through logarithmic ratio transformation, extracts elemental combination features through robust principal component analysis, and combines weighted distance-based D-Kmeans++ clustering with spatial proximity and attribute similarity. The organic combination of these three methods produces a synergistic effect: ILR transformation makes the data more suitable for Euclidean distance calculation; RPCA dimensionality reduction retains the main components that reflect the symbiotic combination patterns of ore-forming elements, reducing the input dimension for clustering; and weighted distance D-Kmeans++ introduces the coefficient of variation weight during clustering, giving high-variance elements (usually more closely related to ore formation) a greater proportion in distance calculation. These three methods are not simply a superposition of techniques, but rather mutually reinforcing, ultimately resulting in significantly higher accuracy in determining the lower limit of anomalies than either step alone or any combination of two. Comparative experiments demonstrate that, compared to traditional K-means++ and fractal methods, the method in this application improves the anomaly hit rate by more than 20% and reduces the anomaly area by more than 30%.
Smart Images
Figure CN122923115A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of big data geoscience data processing, and more specifically, relates to a method and system for determining the lower limit of geochemical anomalies. Background Technology
[0002] Geochemical prospecting methods involve analyzing geochemical data to identify anomalies related to mineralization. Determining the lower limit of anomalies is a crucial step in geochemical prospecting; it serves as a critical threshold distinguishing geological background from mineralization anomalies, directly impacting the accuracy of anomaly identification and mineral resource location.
[0003] However, existing methods for determining the lower limit of geochemical anomalies mainly suffer from the following technical bottlenecks: First, geochemical data exhibits closure, meaning the sum of the abundances of all elements is a fixed constant. Traditional data processing methods (such as principal component analysis) cannot directly handle this type of component data, leading to "spurious correlations" that distort the true symbiotic relationships between elements and result in inaccurate anomaly identification.
[0004] Secondly, conventional principal component analysis is highly susceptible to outliers (i.e., mineralized samples). When the data contains samples with high mineralization content, the variance structure of the principal components will be dominated by a few outliers, causing the extracted elemental combinations to fail to represent the true situation of the background geological body and reducing the accuracy of anomaly screening.
[0005] Furthermore, traditional clustering methods (such as K-means) classify samples based solely on element content values, ignoring the spatial location information of the sampling points. This results in a lack of spatial continuity in the clustering results, often grouping spatially distant points with similar element contents into one category, thus failing to form geologically significant, spatially contiguous anomalous areas.
[0006] Finally, traditional methods often treat different elements or oxides as equally weighted features. In geochemical prospecting, the coefficients of variation of elements closely related to mineralization are usually large, and their content variations have higher indicative value for identifying mineralization anomalies. However, equal weighting weakens the contribution of these key elements and reduces the sensitivity of anomaly identification.
[0007] Therefore, there is an urgent need for a method to determine the lower limit of geochemical anomalies that can simultaneously solve the problems of data closure effect, outlier interference, missing spatial features, and unreasonable feature weight allocation. Summary of the Invention
[0008] In view of the shortcomings of the prior art, the purpose of this application is to provide a method and system for determining the lower limit of geochemical anomalies, which can accurately and reliably determine the lower limit of geochemical anomalies.
[0009] To achieve the above objectives, in a first aspect, this application provides a method for determining the lower limit of geochemical anomalies, comprising the following steps: S10: Obtain geological and geochemical data of the study area, preprocess the geological data to obtain geological data containing the spatial coordinates of each sampling point, construct a geochemical dataset containing multiple element content values based on the geochemical data, and obtain the coefficient of variation of each element content value. S20, Perform an equal-interval logarithmic ratio transformation on the element content values in the geochemical dataset to obtain the transformed geochemical data; S30, Perform robust principal component analysis on the transformed geochemical data to obtain the combination characteristics of elements; S40, the spatial coordinates in the geological data are combined with the transformed geochemical data. The weight of each element content value is determined according to the coefficient of variation. The lower limit of anomalies for the elements in the study area is determined by the weighted distance-based D-Kmeans++ clustering method. The weighted distance is calculated based on the spatial coordinate difference from each sampling point to the cluster center, the difference in content value of each element, and their respective weights. The equidistant logarithmic ratio transformation, robust principal component analysis, and weighted distance-based D-Kmeans++ clustering work together to ensure that the clustering results simultaneously reflect the element combination characteristics, spatial proximity, and data dispersion, thereby synergistically improving the accuracy of the lower limit of anomalies. A single-element anomaly distribution map is drawn based on the lower limit of anomalies and the combination characteristics of the elements.
[0010] As a further preferred embodiment, in step S10, the coefficient of variation of the content value of each element is... CV The calculation formula is:
[0011] In the formula, S Standard deviation of the content values for each element; The average value of the content of each element.
[0012] As a further preferred embodiment, in step S20, the equidistant logarithmic ratio transformation adopts the ilr(x) transformation, where the relative distance between the transformed data and the original data variables remains unchanged. The transformation formula is:
[0013] In the formula, D The number of dimensions in the component data space; For the first i The content value of each element; For the first j The content value of each element.
[0014] As a further preferred embodiment, in step S30, the robust principal component analysis includes: extracting the first principal component, the second principal component, and the third principal component; combining the first principal component, the second principal component, and the third principal component extracted by the robust principal component analysis in pairs as the horizontal and vertical axes to draw a bipolar plot; drawing a perpendicular line from the origin to the horizontal and vertical axes, and dividing each component in the bipolar plot into two parts with positive and negative loads, with the 0 axis as the boundary, to obtain the combination features of the elements.
[0015] As a further preferred embodiment, in step S40, the D-Kmeans++ clustering method considers the spatial coordinates of each sampling point and the element content value of each sampling point in a unified manner, so that sampling points of the same class are similar in element content and spatially adjacent; when calculating the distance, different weights are assigned to the spatial coordinates X, Y and each element content value, and the weight of each element content value is determined according to the coefficient of variation of the element content value.
[0016] As a further preferred option, the weight of the spatial coordinate X is denoted as... The weight of the spatial coordinate Y is denoted as , No. j Weight of each element content value The calculation formula is:
[0017] In the formula, For the first j The coefficient of variation of the content values of each element; n The total number of element content values; For the first k The coefficient of variation of the content values of each element; Distance from sampling point to cluster center The calculation formula is:
[0018] In the formula, , For the first i Spatial coordinates of each sampling point; , For the first j Spatial coordinates of the cluster centers; For the first i The sampling point of the first sampling point k The content value of each element; For the first j The first cluster center k The content value of each element; , and They represent X , Y and nThe weights of each attribute satisfy... .
[0019] As a further preferred embodiment, in step S40, the D-Kmeans++ clustering method includes: S41, randomly select one point from all sampling points as the first initial cluster center; S42, For each sampling point, calculate the shortest distance from the sampling point to the known cluster center, denoted as D(x); S43, calculate the probability that a sampling point is selected as the next cluster center based on D(x) for each sampling point. , The next cluster center is selected using the roulette wheel selection method. S44, Repeat steps S42 and S43 to select K-1 cluster centers; after each new cluster center is determined, re-evaluate the distance of all sampling points to all current cluster centers and update the distance of each sampling point to the nearest cluster center; S45, classify the remaining sampling points and assign them to the nearest corresponding cluster based on their weighted distance to the K cluster centers; S46, Statistically analyze the sampling points within each cluster, calculate the center point of the cluster, and use it as the updated cluster center; S47, assess whether the cluster centers have changed. If they have changed, proceed to steps S45 and S46; otherwise, proceed to the next step. S48 yields the final K cluster centers and the category to which each sampling point belongs.
[0020] Secondly, this application provides a system for determining the lower limit of geochemical anomalies, used to implement the method for determining the lower limit of geochemical anomalies as described in any one of the above statements, comprising: The data acquisition module is used to acquire geological and geochemical data of the study area, preprocess the geological data to obtain geological data containing the spatial coordinates of each sampling point, construct a geochemical dataset containing multiple element content values based on the geochemical data, and obtain the coefficient of variation of each element content value. The transformation module is used to perform an equidistant logarithmic ratio transformation on the element content values in the geochemical dataset to obtain the transformed geochemical data. The analysis module is used to perform robust principal component analysis on the transformed geochemical data to obtain the combination characteristics of elements; The clustering module is used to combine the spatial coordinates in the geological data with the transformed geochemical data, determine the weight of each element content value according to the coefficient of variation, and use the weighted distance-based D-Kmeans++ clustering method to determine the lower limit of anomalies of elements in the study area. The weighted distance is calculated based on the spatial coordinate difference between sampling points, the difference of each element content value, and their respective weights. Then, a single-element anomaly distribution map is drawn based on the lower limit of anomalies.
[0021] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for determining the lower limit of geochemical anomalies as described in any one of the above descriptions.
[0022] Fourthly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for determining the lower limit of geochemical anomalies as described in any of the above-mentioned methods.
[0023] The method for determining the lower limit of geochemical anomalies provided in this application has the following advantages: It eliminates closure effects through logarithmic ratio transformation, extracts elemental combination features through robust principal component analysis, and combines weighted distance-based D-Kmeans++ clustering with spatial proximity and attribute similarity. The organic combination of these three methods produces a synergistic effect: ILR transformation makes the data more suitable for Euclidean distance calculation; RPCA dimensionality reduction retains the main components that reflect the symbiotic combination patterns of ore-forming elements, reducing the input dimension for clustering; and weighted distance D-Kmeans++ introduces the coefficient of variation weight during clustering, giving high-variance elements (usually more closely related to ore formation) a greater proportion in distance calculation. These three methods are not simply a superposition of techniques, but rather mutually reinforcing, ultimately resulting in significantly higher accuracy in determining the lower limit of anomalies than either step alone or any combination of two. Comparative experiments demonstrate that, compared to traditional K-means++ and fractal methods, the method in this application improves the anomaly hit rate by more than 20% and reduces the anomaly area by more than 30%. Attached Figure Description
[0024] Figure 1 This is a flowchart of the D-KMeans++ algorithm provided in the embodiments of this application; Figure 2 This is a distribution map of ILR transform data provided in an embodiment of this application; Figure 3 These are the bi-labeled graphs of various data provided in the embodiments of this application; wherein, (a) is the PC1 / PC2 bi-labeled graph, (b) is the PC1 / PC3 bi-labeled graph, and (c) is the PC2 / PC3 bi-labeled graph; Figure 4This is the Ag element D-Kmeans++ cluster point diagram provided in the embodiments of this application; Figure 5 This is a D-Kmeans++ clustering point diagram of Pb elements provided in the embodiments of this application; Figure 6 This is a D-Kmeans++ clustering point diagram of Zn elements provided in the embodiments of this application. Detailed Implementation
[0025] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0026] This application provides a method for determining the lower limit of geochemical anomalies based on the D-Kmeans++ algorithm. It aims to perform statistical analysis, logarithmic ratio and robust principal component analysis, and D-Kmeans++ clustering on geochemical data to determine the lower limit value of geochemical anomalies, thereby improving the accuracy and efficiency of geochemical mineral exploration.
[0027] Specifically, the method for determining the lower limit of geochemical anomalies provided in this application includes steps S10 to S40, which are detailed below: Step S10 involves acquiring geological and geochemical data for the study area. The geological data is preprocessed to obtain geological data containing the spatial coordinates of each sampling point. A geochemical dataset containing multiple elemental content values is constructed based on the geochemical data, and the coefficient of variation for each elemental content value is obtained. This step allows for the organization and statistical analysis of the original geological and geochemical data. The coefficient of variation measures the dispersion of the elemental content distribution, providing a basis for subsequent weight allocation.
[0028] Among them, the coefficient of variation of the content value of each element CV The calculation formula is:
[0029] In the formula, S Standard deviation of the content values for each element; The average value of the content of each element.
[0030] It's important to note that the coefficient of variation (CV), the ratio of standard deviation to mean, eliminates the influence of dimensions and reflects the relative dispersion of the data. In geochemistry, elements with high CVs typically exhibit stronger spatial variability and are often closely related to later hydrothermal activity and mineralization, serving as indicator elements for locating concealed ore bodies. Therefore, assigning greater weight to elements with high CVs allows cluster analysis to focus more on the differences in the content of ore-forming elements while reducing the interference from low-variance (background) elements. This differs from the traditional K-means approach of giving equal weight to all features; it represents a weighting strategy guided by explicit prior geochemical knowledge.
[0031] Step S20: Perform an equidistant logarithmic ratio transformation on the elemental content values in the geochemical dataset to obtain the transformed geochemical data.
[0032] In this step, the ILR transformation can be used to eliminate the closure effect. That is, when the data consists of multiple non-negative components, and the sum of these components is a fixed constant, the sum is forcibly constrained. An increase in any component will inevitably lead to a decrease in at least one other component, thus producing spurious correlations or distorted distance relationships in statistical analysis. The relative distance between the data after the logarithmic ratio transformation and the data before the transformation remains unchanged. The calculation formula is as follows:
[0033] In the formula, D is the number of dimensions in the component data space; For the first i The content value of each element; For the first j The content value of each element.
[0034] Step S30: Perform robust principal component analysis on the transformed geochemical data to obtain the elemental combination characteristics, thereby reducing the impact of outliers on the analysis results. Pairwise, plot biplots are created using principal components 1 (PC1), 2 (PC2), and 3 (PC3). Each component in the biplot is divided into two groups, and the elemental combination characteristics of each component can be interpreted from the corresponding biplot: With the origin as the center, draw perpendicular lines to the horizontal and vertical coordinate axes, using the 0-axis as the boundary, dividing the data into positive and negative parts, thus obtaining the elemental combination characteristics.
[0035] It should be noted that geochemical data is typical compositional data, where the sum of the content of each element is a constant (e.g., 100% or 10^6 ppm). Directly performing PCA or RPCA will produce spurious correlations. The logarithmic ratio transformation can map compositional data from simplex space to real space, eliminating closure effects and making the transformed data suitable for Euclidean distance calculations and linear analysis. Performing RPCA after the ILR transformation can both eliminate closure effects and resist the influence of outliers; the combination of these two methods is the standard procedure for multivariate statistical analysis of compositional data. This application combines this standard procedure with weighted D-Kmeans++ clustering for the first time, forming a complete chain for determining the lower bound of outliers.
[0036] Step S40 combines the spatial coordinates from the geological data with the transformed geochemical data, determines the weight of each element's content value based on the coefficient of variation, and uses a weighted distance-based D-Kmeans++ clustering method to determine the anomaly lower limit for elements in the study area. The weighted distance is calculated based on the spatial coordinate difference from each sampling point to the cluster center, the difference in each element's content value, and its corresponding weight. The equidistant logarithmic ratio transformation, robust principal component analysis, and weighted distance-based D-Kmeans++ clustering work together to ensure that the clustering results simultaneously reflect element combination characteristics, spatial proximity, and data dispersion, synergistically improving the accuracy of anomaly lower limit determination. A single-element anomaly distribution map is then drawn based on the anomaly lower limit and the element combination characteristics. This approach ensures that samples of the same class are similar in attributes and spatially adjacent, combining the spatial and attribute characteristics of geochemical data to make the clustering results more accurate and reliable. It overcomes the drawbacks of using a single anomaly lower limit for anomaly extraction across the entire study area due to its large size and complex geological conditions.
[0037] In this step, the D-Kmeans++ clustering method considers spatial coordinates and sample attribute features in a unified manner, which makes samples of the same class similar in attributes and spatially adjacent. Combining the spatial and attribute features of geochemical data can make the clustering results more accurate and reliable.
[0038] In this context, spatial coordinates X and Y, as well as each attribute feature, are assigned different weights, with X having a weight of 1. The weight of Y is The calculation formula is as follows:
[0039] In the formula, For the first j The coefficient of variation of the content values of each element; n The total number of element content values; For the first kThe coefficient of variation of the content values of each element.
[0040] The weights of the n attribute features are ,distance The calculation formula is as follows:
[0041] In the formula, , For the first i Spatial coordinates of each sampling point; , For the first j Spatial coordinates of the cluster centers; For the first i The sampling point of the first sampling point k The content value of each element; For the first j The first cluster center k The content value of each element; X , Y and n The weights of the attributes satisfy On the other hand, the selection of spatial coordinate weights should match the scale of the study area: when the study area is large (e.g., >10000 km²), spatial coordinates have a greater impact on the clustering results, and w can be appropriately increased. X w Y The sum of the coordinate weights should be between 0.5 and 0.8. When the study area is small (e.g., <1000 km²), spatial proximity is relatively unimportant, and the sum of the coordinate weights can be reduced to 0.1 to 0.3. Preferably, in this application, the coordinate weights of Ag, Pb, and Zn elements are all set to 0.125 (half for latitude and half for longitude), and the sum of the attribute weights is 0.75, which yields the best results. In practical applications, a grid search method can also be used to optimize the weight combination. Overall, this distance calculation formula assigns different weights to various attributes of the study area. By reasonably selecting and adjusting the weights, more accurate and comprehensive analysis results can be provided.
[0042] Specifically, such as Figure 1 As shown, the D-Kmeans++ algorithm provided in this application is as follows: (1) Randomly select a point from m samples and designate it as the initial cluster center c1.
[0043] (2) For each data point, calculate its shortest distance to the known cluster center, denoted as . .
[0044] (3) For each data point x, the probability of that data point being selected as the next cluster center is obtained according to the formula. , The next cluster center is randomly selected from all data points using a roulette wheel selection method.
[0045] (4) Repeat steps (2) and (3) to select K-1 cluster centers. After each new cluster center is determined, re-evaluate the distance of all points to all current cluster centers and update the distance of each data point to the nearest cluster center.
[0046] (5) Classify the remaining data points and assign them to the nearest corresponding cluster based on their distance to the K cluster centers.
[0047] (6) Perform statistics on the data within each cluster, calculate the cluster center point, and use it as the updated cluster center.
[0048] (7) Assess whether the cluster centers have changed. If they have changed, proceed with steps (5) and (6); otherwise, proceed to the next step.
[0049] (8) Once the final K cluster centers and the category to which each sample belongs are obtained, the algorithm ends.
[0050] The method for determining the lower limit of geochemical anomalies provided in this application has the following effects: by using logarithmic ratio transformation, the closure effect in geochemical data can be eliminated; by using robust principal component analysis, the influence of outliers on the analysis results can be reduced; by using the weighted distance-based D-Kmeans++ clustering method, the spatial and attribute characteristics of geochemical data can be combined, taking into account both the attribute similarity of sampling points and spatial proximity, ensuring that the cluster centers are evenly distributed in space, thereby avoiding the influence of local extrema on the clustering results and making the clustering results more accurate and reliable.
[0051] The following is a specific implementation example of this application: Step 1: A total of 6749 samples were collected from the study area at a density of 2 km × 2 km. Each sample contained 32 elements, including Ag, As, and Au, and 7 oxides, including Al₂O₃, CaO, Fe₂O₃, K₂O, MgO, Na₂O, and SiO₂. Statistical analysis of the geochemical data from the study area yielded the minimum, maximum, average, standard deviation, coefficient of variation, and enrichment coefficient for each element (or oxide), as shown in Table 1.
[0052] Table 1. Characteristics of elemental content in the study area
[0053] Step 2 involves analyzing Ag, As, Bi, Cd, Hg, Mn, Mo, Pb, Sb, W, Zn, and Fe2O3 in 6749 samples within the region, performing ILR transformation. The transformed data exhibits a more concentrated distribution compared to the original data, effectively eliminating the order-of-magnitude differences in element content. Figure 2 As shown.
[0054] Step 3: The new variables after ILR transformation are difficult to interpret, greatly limiting the application of this method. RPCA analysis can be performed on the ILR-transformed data. PCA scores and loadings can be transformed to CLR space using orthogonal bases, making the data easier to interpret and understand. The results of principal component analysis can be displayed using a biplot. In the biplot of principal components after ILR transformation, the 12 elements are more evenly distributed, exhibiting a radial distribution, effectively eliminating the closure effect present in the component data. Figure 3 As shown.
[0055] Step 4: Taking Ag, Pb, and Zn as examples, the weight settings can be found in Table 2. The D-Kmeans++ algorithm is executed for all possible weight combinations to obtain the weight combination that best classifies each element.
[0056] Table 2 Attribute Weight Table
[0057] The D-Kmeans++ algorithm was used to cluster data based on Ag element content, geological unit, and latitude and longitude coordinates. The results show that (e.g.) Figure 4 As shown in the figure, the best clustering effect was achieved when the dataset was divided into five categories. The distribution of category I data is basically consistent with the distribution of Triassic strata in the study area, category II data has a good match with Cambrian strata in the study area, category III data is mainly distributed in Permian strata, and categories IV and V data are consistent with the distribution of Carboniferous strata in the region. This indicates that the clustering results are closely related to the geological unit and spatial location of the samples.
[0058] Clustering of Pb element content, geological unit, and latitude / longitude coordinates using the D-Kmeans++ algorithm revealed that the best clustering effect was achieved when the dataset was divided into six classes (e.g., ...). Figure 5 (As shown). Analysis reveals that Class I data is dominated by low background data, primarily located in the Triassic strata; Class III, IV, and V data have higher mean values. The distribution of Class VI and V data shows good agreement with the high-value distribution areas in the Pb element content distribution map of the study area, while the distribution of Class I data matches the low-value distribution of Pb, indicating that the D-Kmeans++ algorithm can effectively cluster the data in the study area.
[0059] Clustering of Zn element data using the D-Kmeans++ algorithm showed that the best clustering effect was achieved when the dataset was divided into five classes (e.g., Figure 6(As shown). Comprehensive analysis of the Zn elemental content distribution map and clustering results revealed that the clustering results were primarily influenced by elemental content values. Category I data were mainly distributed in low background areas, accounting for approximately 88% of the total sample size in the study area. Categories IV and V contained data with higher mean values, primarily located in the Carboniferous and Permian strata. The results indicate that the D-Kmeans++ clustering results show good agreement with the stratigraphy and Zn elemental content distribution in the study area, demonstrating the effectiveness of the clustering.
[0060] To verify the effectiveness of the weighted distance D-Kmeans++ clustering proposed in this application, the same dataset (6749 samples from the study area, Ag, Pb, and Zn elements) was used, and the following three methods were employed to determine the lower bound of anomalies: Traditional K-means clustering (using Euclidean distance, without considering spatial coordinates and the weight of the coefficient of variation); Standard K-means++ clustering (using Euclidean distance, only optimizing the initial center selection); The weighted distance D-Kmeans++ clustering proposed in this application (with weights set according to Table 2)
[0061] Evaluation indicators: Using known mineral occurrences (or known mineralized alteration zones) in the study area as the verification standard, the hit rate of the anomalous areas delineated by each method to known mineral occurrences (number of known mineral occurrences contained in the anomalous area / total number of mineral occurrences) and the anomalous area ratio (area of the anomalous area / total area of the study area; the smaller the anomalous area and the higher the hit rate, the better the effect).
[0062] The experimental results are shown in Table 3.
[0063] Table 3 Experimental Results (Example)
[0064] The results show that the proposed method improves the hit rate by an average of about 20 percentage points and reduces the proportion of anomalous areas by about one-third, indicating that the weighted distance D-Kmeans++ can more accurately identify anomalous areas related to mineralization and reduce the interference of false anomalies. This progress is due to the synergistic effect of using the coefficient of variation as a weight while taking into account both spatial proximity and attribute similarity, rather than a simple summation of the steps.
[0065] Further comparison: The method of this application is compared with the classic fractal method (concentration-area fractal method) for determining the lower limit of geochemical anomalies. For Pb, the fractal method delineates an anomaly area of 14% with a hit rate of 71%; the method of this application has a hit rate of 82% and an anomaly area of 10%, which is better than the fractal method.
[0066] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.
Claims
1. A method for determining the lower limit of geochemical anomalies, characterized in that, Includes the following steps: S10: Obtain geological and geochemical data of the study area, preprocess the geological data to obtain geological data containing the spatial coordinates of each sampling point, construct a geochemical dataset containing multiple element content values based on the geochemical data, and obtain the coefficient of variation of each element content value. S20, Perform an equal-interval logarithmic ratio transformation on the element content values in the geochemical dataset to obtain the transformed geochemical data; S30, Perform robust principal component analysis on the transformed geochemical data to obtain the combination characteristics of elements; S40, the spatial coordinates in the geological data are combined with the transformed geochemical data. The weight of each element content value is determined according to the coefficient of variation. The lower limit of anomalies for the elements in the study area is determined by the weighted distance-based D-Kmeans++ clustering method. The weighted distance is calculated based on the spatial coordinate difference from each sampling point to the cluster center, the difference in content value of each element, and their respective weights. The equidistant logarithmic ratio transformation, robust principal component analysis, and weighted distance-based D-Kmeans++ clustering work together to ensure that the clustering results simultaneously reflect the element combination characteristics, spatial proximity, and data dispersion, thereby synergistically improving the accuracy of the lower limit of anomalies. A single-element anomaly distribution map is drawn based on the lower limit of anomalies and the combination characteristics of the elements.
2. The method for determining the lower limit of geochemical anomalies as described in claim 1, characterized in that, In step S10, the coefficient of variation of the content value of each element CV The calculation formula is: In the formula, S The standard deviation of the content values for each element; The average value of the content of each element.
3. The method for determining the lower limit of geochemical anomalies as described in claim 1, characterized in that, In step S20, the logarithmic ratio transformation uses the ilr(x) transformation, where the relative distance between the transformed data and the original data variables remains unchanged. The transformation formula is: In the formula, D The number of dimensions in the component data space; For the first i The content value of each element; For the first j The content value of each element.
4. The method for determining the lower limit of geochemical anomalies as described in claim 1, characterized in that, In step S30, the robust principal component analysis includes: extracting the first principal component, the second principal component, and the third principal component; combining the first principal component, the second principal component, and the third principal component extracted by the robust principal component analysis in pairs as the horizontal and vertical axes to draw a bipolar plot; drawing a perpendicular line from the origin to the horizontal and vertical axes, and dividing each component in the bipolar plot into two parts with positive and negative loads, with the 0 axis as the boundary, to obtain the combination features of the elements.
5. The method for determining the lower limit of geochemical anomalies as described in claim 1, characterized in that, In step S40, the D-Kmeans++ clustering method considers the spatial coordinates of each sampling point and the element content value of each sampling point in a unified manner, so that sampling points of the same class are similar in element content and spatially adjacent; when calculating the distance, different weights are assigned to the spatial coordinates X, Y and each element content value, and the weight of each element content value is determined according to the coefficient of variation of the element content value.
6. The method for determining the lower limit of geochemical anomalies as described in claim 5, characterized in that, The weight of the spatial coordinate X is denoted as The weight of the spatial coordinate Y is denoted as , No. j Weight of each element content value The calculation formula is: In the formula, For the first j The coefficient of variation of the content values of each element; n The total number of element content values; For the first k The coefficient of variation of the content values of each element; Distance from sampling point to cluster center The calculation formula is: In the formula, , For the first i Spatial coordinates of each sampling point; , For the first j Spatial coordinates of the cluster centers; For the first i The sampling point of the first sampling point k The content value of each element; For the first j The first cluster center k The content value of each element; , and They represent X , Y and n The weights of each attribute satisfy... .
7. The method for determining the lower limit of geochemical anomalies as described in claim 6, characterized in that, In step S40, the D-Kmeans++ clustering method includes: S41, randomly select one point from all sampling points as the first initial cluster center; S42, For each sampling point, calculate the shortest distance from the sampling point to the known cluster center, denoted as D(x); S43, calculate the probability that a sampling point is selected as the next cluster center based on D(x) for each sampling point. , The next cluster center is selected using the roulette wheel selection method. S44, Repeat steps S42 and S43 to select K-1 cluster centers; after each new cluster center is determined, re-evaluate the distance of all sampling points to all current cluster centers and update the distance of each sampling point to the nearest cluster center; S45, classify the remaining sampling points and assign them to the nearest corresponding cluster based on their weighted distance to the K cluster centers; S46, Statistically analyze the sampling points within each cluster, calculate the center point of the cluster, and use it as the updated cluster center; S47, assess whether the cluster centers have changed. If they have changed, proceed to steps S45 and S46; otherwise, proceed to the next step. S48 yields the final K cluster centers and the category to which each sampling point belongs.
8. A system for determining the lower limit of geochemical anomalies, characterized in that, A method for determining the lower limit of geochemical anomalies as described in any one of claims 1 to 7, comprising: The data acquisition module is used to acquire geological and geochemical data of the study area, preprocess the geological data to obtain geological data containing the spatial coordinates of each sampling point, construct a geochemical dataset containing multiple element content values based on the geochemical data, and obtain the coefficient of variation of each element content value. The transformation module is used to perform an equidistant logarithmic ratio transformation on the element content values in the geochemical dataset to obtain the transformed geochemical data. The analysis module is used to perform robust principal component analysis on the transformed geochemical data to obtain the combination characteristics of elements; The clustering module is used to combine the spatial coordinates in the geological data with the transformed geochemical data, determine the weight of each element content value according to the coefficient of variation, and use the weighted distance-based D-Kmeans++ clustering method to determine the lower limit of anomalies of elements in the study area. The weighted distance is calculated based on the spatial coordinate difference between sampling points, the difference of each element content value, and their respective weights. Then, a single-element anomaly distribution map is drawn based on the lower limit of anomalies.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for determining the lower limit of geochemical anomalies as described in any one of claims 1 to 7.
10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method for determining the lower limit of geochemical anomalies as described in any one of claims 1 to 7.