An interpretable control factor identification method for land subsidence sensitivity
By combining multi-temporal synthetic aperture radar interferometry and machine learning models, the interpretability problem of machine learning models in land subsidence simulation was solved, the contribution of influencing factors to land subsidence and spatial heterogeneity were quantified, and a scientific basis for regulation was provided.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NANTONG UNIV
- Filing Date
- 2026-02-06
- Publication Date
- 2026-04-14
AI Technical Summary
Existing machine learning models have poor interpretability in land subsidence simulations, making it difficult to reveal the nonlinear response characteristics between influencing factors and regional subsidence, and lacking an understanding of the mechanisms underlying the spatial heterogeneity of land subsidence.
Surface deformation information was acquired using multi-temporal synthetic aperture radar interferometry (MT-InSAR). The optimal grid size was determined by combining Moran's I coefficient difference. XGBoost and SHAP models were constructed, and multicollinearity analysis was performed to quantify the marginal contribution and spatial heterogeneity of influencing factors to land subsidence.
This study enabled the interpretable identification of factors controlling land subsidence sensitivity, quantified the marginal contribution of influencing factors to regional land subsidence, revealed spatial heterogeneity, and provided a theoretical basis for the scientific regulation of land subsidence.
Smart Images
Figure CN121660133B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of geological disaster prevention technology, and in particular to an interpretable method for identifying control factors of land subsidence sensitivity. Background Technology
[0002] Land subsidence is a phenomenon caused by the combined effects of natural factors and human activities, resulting in a loss of land elevation. It increases the risk of urban flooding, seawater intrusion, and in severe cases, can induce a series of environmental disasters such as ground collapse and ground fissures, forming a disaster chain.
[0003] Simulation and prediction of land subsidence processes can provide scientific support for regional subsidence prevention and control. Machine learning intelligent models developed using big data and Geo-AI technologies are not limited by complex physical parameters such as regional hydrogeology, and can better solve the problem of low efficiency of traditional mathematical and statistical methods. However, due to the black-box nature of machine learning models, it is difficult to reveal the nonlinear response characteristics between influencing factors and regional subsidence, quantify the marginal contribution of influencing factors to land subsidence, and understand the causal mechanism of spatial heterogeneity of land subsidence.
[0004] How to solve the above-mentioned technical problems is the challenge facing this invention. Summary of the Invention
[0005] The purpose of this invention is to provide an interpretable method for identifying control factors of land subsidence sensitivity. This method solves the technical problem that existing machine learning models have poor interpretability and fail to clearly reveal the influence of various factors on regional subsidence. It quantifies the marginal contribution of influencing factors to regional land subsidence, reveals the spatial heterogeneity of regional subsidence, and provides a certain theoretical basis for the scientific regulation of land subsidence. This invention reveals and quantifies the influence weight of environmental factors on land subsidence and the spatial heterogeneity of their effects, further deepening the understanding of the causes and mechanisms of differential land subsidence and providing a basis for the prevention and control of regional land subsidence.
[0006] To achieve the aforementioned objectives, the present invention employs the following technical solution: a method for identifying interpretable ground subsidence sensitivity control factors, comprising the following steps:
[0007] Step S1: Acquisition of regional surface deformation information
[0008] Multi-track radar images covering the study area were selected, and the line-of-sight deformation information of each track radar image was obtained using the multi-temporal synthetic aperture radar interferometry (MT-InSAR) method.
[0009] After converting the line-of-sight deformation information into vertical deformation results, leveling data was used to verify the accuracy of the multi-temporal synthetic aperture radar interferometry (MT-InSAR) results.
[0010] Based on ensuring the accuracy of the monitoring results of multi-temporal synthetic aperture radar interferometry (MT-InSAR), the least squares method is used to calculate the offset between the overlapping areas of adjacent orbits, and the MT-InSAR results of multi-track images are fused to obtain the regional surface deformation results.
[0011] Step S2: Obtaining the ground settlement sensitivity assessment unit
[0012] Calculate Moran's I coefficient for the original surface deformation data and the resampled data, respectively;
[0013] The Moran's I coefficient difference between the original surface deformation data and the resampled data with different grid sizes is calculated. The grid size corresponding to the smallest Moran's I coefficient difference is taken as the optimal spatial grid size for surface deformation data resampling, so as to preserve the spatial autocorrelation of the original data to the maximum extent.
[0014] The surface deformation data were resampled according to the above-mentioned optimal grid size, and the resampled data was used as the evaluation unit for subsequent ground subsidence sensitivity analysis.
[0015] Step 3: Obtaining factors affecting ground subsidence
[0016] Nine factors were selected to construct a dataset of regional land subsidence influencing factors: elevation, aspect, slope, distance from fault (DFF), distance from river (DFR), fault density (FD), river network density (RD), deep groundwater level (DGL), and shallow groundwater level (SGL).
[0017] Multicollinearity analysis was performed on the above-mentioned influencing factors using variance inflation factor (VIF) and tolerance (TOL) to avoid the impact of high correlation between the selected factors on the subsequent land subsidence simulation results.
[0018] Using QGIS, a mapping relationship was established between the influencing factors obtained through multicollinearity analysis and the surface deformation data obtained from the above resampling.
[0019] Step S4: Construction of an interpretable ground settlement sensitivity assessment model and identification of controlling factors
[0020] Based on the surface deformation and environmental factor datasets in steps S2 and S3, an interpretable land subsidence sensitivity assessment model is constructed by combining XGBoost and SHAP models. The regional land subsidence sensitivity zoning results are obtained, and the influence of influencing factors on regional land subsidence and the spatial heterogeneity of their effects are quantified.
[0021] Preferably, step S4 specifically includes the following sub-steps.
[0022] The construction of the interpretable ground subsidence sensitivity assessment model described in step S4 specifically includes the following steps:
[0023] Using the multicollinearity analysis in step S3 as the model input data and the resampled surface deformation information in step S2 as the model output label, the XGBoost regression model is used to construct a regional land subsidence sensitivity assessment model.
[0024] During the training process of the ground subsidence sensitivity assessment model, the data is randomly divided into a training set and a test set, with the training set accounting for 80% of the entire dataset.
[0025] Construct a pre-trained network and use Bayesian-optimized 5-fold cross-validation to tune hyperparameters and determine the optimal configuration;
[0026] The trained model is determined as the land subsidence sensitivity assessment model. The natural discontinuity method is used to classify and zon the model simulation results, and the land subsidence sensitivity of the region is divided into four levels: low, low, high, and high, to obtain the land subsidence sensitivity zoning results of the region.
[0027] By integrating the SHAP model, attribution analysis is performed on the land subsidence model to obtain the marginal contribution and nonlinear effect of environmental factors on land subsidence in the region.
[0028] The identification of ground subsidence sensitivity control factors in step S4 is as follows:
[0029] A SHAP attribution analysis was conducted on the ground subsidence model. By calculating the SHAP values of the influencing factors in the XGBoost model, the contribution of each factor to surface deformation was quantified, key influencing factors of ground subsidence in the region were identified, and the marginal contribution and threshold effect of key influencing factors on subsidence were revealed.
[0030] Based on this, and combined with QGIS spatial analysis software, the SHAP values of key influencing factors are mapped onto space, serving as a quantitative analysis result of the spatial heterogeneity of the impact of influencing factors on land subsidence in the region.
[0031] Meanwhile, the present invention proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed, it implements the steps of the method described in the present invention.
[0032] Furthermore, the present invention proposes a computer-readable storage medium having a computer program stored thereon, the computer program being configured to implement the steps of the method described in the present invention when invoked by a processor.
[0033] Finally, the present invention proposes a computer program product, including a computer program / instructions, characterized in that the computer program / instructions, when executed by a processor, implement the steps of the method described in the present invention.
[0034] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0035] 1. This invention addresses the technical problem of poor interpretability and difficulty in clarifying the influence of various factors on regional subsidence in existing ground subsidence machine learning models. It proposes an interpretable method for identifying ground subsidence sensitivity control factors, quantitatively analyzes the marginal contribution of influencing factors to regional ground subsidence, and identifies the main influencing factors of regional subsidence.
[0036] 2. In response to the problem that existing machine learning models for land subsidence are unable to reveal the spatial effects of influencing factors on regional subsidence, this invention proposes an interpretable method for identifying control factors of land subsidence sensitivity. By quantifying the spatial nonstationarity of variable relationships, this method reflects the spatial heterogeneity of the effects of influencing factors, thereby further improving the understanding of the causal mechanism of spatial heterogeneity of regional land subsidence. Attached Figure Description
[0037] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.
[0038] Figure 1 This is a flowchart illustrating an interpretable method for identifying control factors of land subsidence sensitivity proposed in this invention.
[0039] Figure 2 This is a map showing the surface deformation data results for the study area after resampling;
[0040] Figure 3 This is a map showing the results of the zoning of ground subsidence sensitivity.
[0041] Figure 4 A schematic diagram illustrating the contribution rate of environmental factors in the land subsidence sensitivity assessment model provided in this embodiment of the invention;
[0042] Figure 5 A schematic diagram of SHAP (Situational Analysis of Surface Deformation Control Factors) provided for embodiments of the present invention;
[0043] Figure 6 A partial dependency graph of key influencing factors provided for embodiments of the present invention;
[0044] Figure 7 A schematic diagram illustrating the spatial heterogeneity of the effects of key influencing factors in embodiments of the present invention. Detailed Implementation
[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. Of course, the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0046] Example 1
[0047] See Figure 1 and Figure 7 This embodiment provides a technical solution: an interpretable method for identifying control factors of land subsidence sensitivity, comprising the following steps:
[0048] S1: Acquisition of regional surface deformation information; It should be noted that, as an example, step S1 is as follows: S11, Select multi-track Sentinel-1A (S1A) images covering the study area, and use the multi-temporal synthetic aperture radar interferometry (MT-InSAR) method to acquire the surface deformation information of the line of sight of each track S1A image.
[0049] S12. After converting the line-of-sight deformation information into vertical deformation results, the accuracy of the MT-InSAR results is verified using leveling data.
[0050] S13. On the basis of ensuring the accuracy of InSAR monitoring results, the least squares method is used to calculate the offset between the overlapping areas of adjacent orbits, and the MT-InSAR results of multi-track images are fused to obtain the regional surface deformation results.
[0051] S2: Obtaining the ground subsidence sensitivity evaluation unit; It should be noted that, as an example, step S2 is as follows: S21, calculate Moran's I coefficients for the original surface deformation data and the data after resampling at different grid sizes respectively.
[0052] S22. Calculate the Moran's I coefficient difference between the original surface deformation data and the resampled data with different grid sizes. The grid size corresponding to the minimum coefficient difference is taken as the optimal spatial grid size for surface deformation data resampling, in order to maximize the preservation of the spatial autocorrelation of the original data. As an example, when the grid size is 4300m, the Moran's I coefficient difference between the resampled data and the original data is minimized.
[0053] S23: Resample the surface deformation data according to the above-mentioned optimal grid size, and use the resampled data as the evaluation unit for subsequent ground subsidence sensitivity analysis. As one embodiment, this invention uses 4300×4300 m as the optimal size to resample the surface deformation results obtained from MT-InSAR monitoring, ultimately obtaining 4088 sampling points. The resampled surface deformation data is as follows: Figure 2As shown.
[0054] S3: Obtaining the influencing factors of land subsidence; It should be noted that, as an example, step S3 is as follows: S31, select nine factors, including elevation, aspect, slope, distance from fault (DFF), distance from river (DFR), fault density (FD), river network density (RD), deep groundwater level (DGL), and shallow groundwater level (SGL), to construct a regional land subsidence influencing factor dataset.
[0055] S32. Multicollinearity analysis was performed on the above-mentioned influencing factors using variance inflation factor (VIF) and tolerance (TOL) to avoid the impact of high correlation among the selected factors on the subsequent land subsidence simulation results. As an example, the multicollinearity analysis results of the selected factors in this invention are shown in Table 1. The calculated VIF for all factors was less than 10, and the TOL was greater than 0.1, indicating that there was no collinearity problem among the factors, and all could be used in the subsequent model construction. S33. A mapping relationship was established between the influencing factors after multicollinearity analysis and the surface deformation data obtained from the above resampling using QGIS. Table 1 shows the results of multicollinearity analysis of the influencing factors:
[0056] Table 1
[0057]
[0058] S4: Construction of an interpretable land subsidence sensitivity assessment model and identification of controlling factors. Before formally describing step S4, it will first be given a general overview for ease of understanding.
[0059] Construction of an interpretable land subsidence sensitivity assessment model and identification of controlling factors. Based on the surface deformation and environmental factor datasets in steps S2 and S3, an interpretable land subsidence sensitivity assessment model is constructed by combining XGBoost and SHAP models. The regional land subsidence sensitivity zoning results are obtained, and the influence of influencing factors on regional land subsidence and the spatial heterogeneity of their effects are quantified.
[0060] Construction of an interpretable land settlement sensitivity assessment model and identification of controlling factors. As one embodiment, step S4 specifically includes the following two aspects: construction of an interpretable land settlement sensitivity assessment model and identification of land settlement sensitivity controlling factors.
[0061] The construction of an interpretable land subsidence sensitivity assessment model in S4 includes the following steps:
[0062] S41. Using the influencing factors obtained through multicollinearity analysis in step S32 as the model input data, and the surface deformation information after resampling in step S2 as the model output label, a regional land subsidence sensitivity assessment model is constructed using the XGBoost regression model.
[0063] S42. During the training process of the ground subsidence sensitivity assessment model, the data is randomly divided into a training set and a test set, wherein the proportion of the training set accounts for 80% of the entire dataset.
[0064] S43. Construct a pre-trained network and use Bayesian-optimized 5-fold cross-validation to tune hyperparameters and determine the optimal configuration. As one embodiment, this invention uses the coefficient of determination (R²). 2 The model performance is evaluated using mean absolute error (MAE), explanatory variance score (EVS), and root mean square error (RMSE). This invention utilizes the R-squared value of the ground subsidence model constructed based on XGBoost. 2 The EVS and MAE values are both 0.96, and the RMSE values are 2.25 mm / yr and 5.4 mm / yr, respectively. All these indicators demonstrate that the model constructed in this invention has high accuracy and reliability.
[0065] S44. The trained model is designated as the land subsidence sensitivity assessment model. The simulation results are then graded and partitioned using the natural discontinuity method, classifying the regional land subsidence sensitivity into four levels: low, low, high, and relatively high. This yields the regional land subsidence sensitivity partitioning results. As an example, the results of the land subsidence sensitivity assessment model constructed in this invention are shown in Table 2, and the regional land subsidence sensitivity partitioning results are as follows: Figure 3 As shown.
[0066] S45. By integrating the SHAP model, attribution analysis is performed on the land subsidence model to obtain the marginal contribution and nonlinear effect of environmental factors on the land subsidence in the region. Table 2 shows the regional land subsidence sensitivity zoning results as follows:
[0067] Table 2
[0068]
[0069] The specific steps for identifying the control factors of land subsidence sensitivity in S4 are as follows:
[0070] A SHAP attribution analysis was conducted on the land subsidence model. By calculating the SHAP values of each influencing factor in the XGBoost model, the contribution of each factor to surface deformation was quantified, key influencing factors of land subsidence in the region were identified, and the marginal contribution and threshold effect of key influencing factors on subsidence were revealed. As one embodiment, this invention ranks and visualizes the characteristic importance of each influencing factor by calculating its SHAP value, and ranks the importance of each factor according to its average absolute SHAP value in the examples. Figure 4 And use a summary diagram to illustrate the impact of each feature on the prediction of the instance model ( Figure 5 Specifically, Figure 5 Each scatter point represents an evaluation unit, and the magnitude of the characteristic value of the influencing factor corresponding to each evaluation unit is shown by the color band on the right. The horizontal axis represents the SHAP value of each evaluation unit. A SHAP value less than 0 indicates that the factor has a negative impact on surface deformation, that is, the probability of ground subsidence is higher; a SHAP value greater than 0 indicates the opposite. Based on this, the main controlling factors of regional ground subsidence sensitivity are screened to reveal their marginal contributions ( Figure 6 ).
[0071] Based on this, and using QGIS spatial analysis software, the SHAP values of key influencing factors are mapped spatially, serving as a quantitative analysis result of the spatial heterogeneity of the influencing factors' impact on land subsidence in the region. As one embodiment, this invention maps the SHAP values of each influencing factor to each evaluation unit, revealing the spatial heterogeneity of the influencing factors' impact on regional subsidence. Figure 7 Specifically, Figure 7 Each point in the figure represents an evaluation unit, and the legend on the right represents the SHAP value of the influencing factor at each evaluation unit. A SHAP value less than 0 indicates that the factor has a negative impact on surface deformation, that is, the greater the probability of ground subsidence; a SHAP value greater than 0 indicates the opposite.
[0072] Example 2: This example proposes an electronic system, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method steps of the present invention.
[0073] Example 3: This example proposes a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements the steps of the method described in this invention, which will not be repeated here.
[0074] Example 4: This example proposes a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, they implement the steps of the method described in this invention, which will not be repeated here.
[0075] It should be noted that the processing flow of embodiments 2-5 corresponds to the specific steps of the method provided in embodiment 1 of the present invention, and has the corresponding functional modules and beneficial effects of the method. Technical details not described in detail in this embodiment can be found in the method provided in embodiment 1 of the present invention.
[0076] The program code used to implement the methods of this application may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.
[0077] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for identifying interpretable control factors of land subsidence sensitivity, characterized in that, The process includes the following steps: Step S1: Acquisition of regional surface deformation information: Multi-temporal synthetic aperture radar interferometry is used to process the multi-track radar images collected in the study area to obtain regional surface deformation information. Step S2: Obtaining ground subsidence sensitivity assessment units: Using Moran's I coefficient difference method, the optimal grid size is selected to resample the surface deformation information obtained in step S1, and it is used as the subsequent ground subsidence sensitivity assessment unit. Step S3: Obtaining ground subsidence influencing factors: Select the following factors to construct a regional ground subsidence influencing factor dataset: elevation, aspect, slope, distance from fault (DFF), distance from river (DFR), fault density (FD), river network density (RD), deep groundwater level (DGL), and shallow groundwater level (SGL). Step S4: Construction of an interpretable land subsidence sensitivity assessment model and identification of controlling factors: Combining the data obtained in steps S2 and S3, an interpretable regional land subsidence sensitivity assessment model is constructed using the Limiting Gradient Boosting (XGBoost) and Shapley additive interpretation (SHAP) models to identify the main controlling factors of regional land subsidence and provide a basis for regional subsidence prevention and control.
2. The method for identifying interpretable land subsidence sensitivity control factors according to claim 1, characterized in that: The acquisition of regional surface deformation information in step S1 includes the following steps: Step S11: Select multi-track radar images covering the study area, and use the multi-temporal synthetic aperture radar interferometry (MT-InSAR) technique to obtain the line-of-sight deformation information of each track radar image. Step S12: After converting the line-of-sight deformation information into vertical deformation results, the accuracy of the multi-temporal synthetic aperture radar interferometry (MT-InSAR) results is verified using leveling data. Step S13: On the basis of ensuring the accuracy of the multi-temporal synthetic aperture radar interferometry (MT-InSAR) monitoring results, the least squares method is used to calculate the offset between the overlapping areas of adjacent orbits, and the MT-InSAR results of multi-track images are fused to obtain the regional surface deformation results.
3. The method for identifying interpretable ground subsidence sensitivity control factors according to claim 2, characterized in that: The acquisition of the ground subsidence sensitivity evaluation unit in step S2 includes the following steps: Step S21: Calculate Moran's I coefficient for the original surface deformation data and the resampled data, respectively; Step S22: Calculate the difference of Moran's I coefficient between the original surface deformation data and the data after resampling at different grid sizes. Take the grid size corresponding to the smallest difference of Moran's I coefficient as the optimal spatial grid size for resampling surface deformation data, so as to preserve the spatial autocorrelation of the original data to the maximum extent. Step S23: Resample the surface deformation data according to the optimal grid size in step S22 above, and use the resampled data as the evaluation unit for subsequent ground subsidence sensitivity analysis.
4. The method for identifying interpretable ground subsidence sensitivity control factors according to claim 3, characterized in that: In step S3, obtaining the ground subsidence influencing factors includes the following steps: Step S31: Select nine factors—elevation, aspect, slope, distance from fault (DFF), distance from river (DFR), fault density (FD), river network density (RD), deep groundwater level (DGL), and shallow groundwater level (SGL)—to construct a dataset of regional land subsidence influencing factors. Step S32: Multicollinearity analysis of the above-mentioned influencing factors is performed using variance expansion factor (VIF) and tolerance (TOL) to avoid the impact of high correlation between the selected factors on the subsequent land subsidence simulation results. Step S33: Use QGIS to establish a mapping relationship between the influencing factors after multicollinearity analysis and the surface deformation data obtained from the above resampling.
5. The method for identifying interpretable land subsidence sensitivity control factors according to claim 4, characterized in that: In step S4, the construction of an interpretable ground subsidence sensitivity assessment model and the identification of controlling factors include the following steps: Based on the surface deformation and environmental factor datasets in steps S2 and S3, an interpretable land subsidence sensitivity assessment model is constructed by combining the Limiting Gradient Boosting (XGBoost) and Shapley additive interpretation (SHAP) models. The results of regional land subsidence sensitivity zoning are obtained, and the influence of influencing factors on regional land subsidence and the spatial heterogeneity of their effects are quantified.
6. The method for identifying interpretable land subsidence sensitivity control factors according to claim 5, characterized in that: The construction of the interpretable ground subsidence sensitivity assessment model in step S4 includes the following steps: Step S41: Using the influencing factors obtained through multicollinearity analysis in step S32 as the model input data, and the resampled surface deformation information in step S2 as the model output label, a regional land subsidence sensitivity assessment model is constructed using the ultimate gradient boosting XGBoost regression model. Step S42: During the training process of the ground subsidence sensitivity assessment model, the data is randomly divided into a training set and a test set, wherein the training set accounts for 80% of the entire dataset; Step S43: Construct a pre-trained network, use Bayesian-optimized 5-fold cross-validation to tune hyperparameters, and determine the optimal configuration; Step S44: The trained model is identified as the land subsidence sensitivity assessment model. The simulation results of the model are classified and zoned using the natural discontinuity method. Areas with a surface deformation rate greater than -20 mm / yr are classified as areas with low land subsidence sensitivity. Areas with a surface deformation rate between -20 mm / yr and -35 mm / yr are classified as areas with low subsidence sensitivity. Areas with a surface deformation rate between -35 mm / yr and -60 mm / yr and less than -60 mm / yr are classified as areas with high subsidence sensitivity and relatively high subsidence sensitivity, respectively. Step S45: Integrate the Shapley additive interpretation SHAP model to perform attribution analysis on the land subsidence model, and obtain the marginal contribution and nonlinear effect of environmental factors on the land subsidence in the region.
7. The method for identifying interpretable land subsidence sensitivity control factors according to claim 6, characterized in that: The identification of ground subsidence sensitivity control factors in step S4 is as follows: A Shapley additive interpretation (SHAP) attribution analysis was conducted on the land subsidence model. By calculating the Shapley additive interpretation (SHAP) values of each influencing factor in the limiting gradient boosting XGBoost model, the contribution of each factor to surface deformation was quantified, key influencing factors of land subsidence in the region were identified, and the marginal contribution and threshold effect of key influencing factors on subsidence were revealed. By combining QGIS spatial analysis software, the Shapley additive interpretation (SHAP) values of key influencing factors are mapped spatially, serving as a quantitative analysis result of the spatial heterogeneity of land subsidence in the region.
8. 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 computer program is executed, it implements the steps of the method as described in any one of claims 1 to 7.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program is configured to implement the steps of the method according to any one of claims 1 to 7 when invoked by a processor.
10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1-7.
Citation Information
Patent Citations
Ground subsidence prediction method and device based on machine learning
CN117470182A
Method for predicting ground surface settlement based on Informer model
CN117910338A