A geological disaster susceptibility evaluation method fusing spatial zoning and factor optimization
By employing spatial partitioning and factor optimization methods, Robust K-means clustering and Fick's law-nearest neighbor model are used to screen feature factors. Combined with particle swarm optimization-random forest model, geological hazard susceptibility prediction is performed, which solves the problems of spatial heterogeneity and factor selection in geological hazard assessment and improves the accuracy and stability of the assessment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- LIAONING TECHNICAL UNIVERSITY
- Filing Date
- 2026-04-09
- Publication Date
- 2026-07-10
AI Technical Summary
Existing geological hazard susceptibility assessment methods fail to effectively reflect the spatial heterogeneity of the geological hazard-generating environment, resulting in unstable assessment results and a lack of adaptability in factor selection, which affects the accuracy of prediction.
The Robust K-means clustering algorithm was used for spatial partitioning, and the study area was divided by combining Thiessen polygon theory. Fick's law-nearest neighbor model was used to screen feature factors, and the particle swarm-random forest model was used to predict the susceptibility of geological disasters. The selection of factors was dynamically adjusted to adapt to the disaster-incubation conditions of different sub-regions.
It improves the stability and reliability of geological hazard susceptibility assessment, reduces interference from redundant factors, ensures spatial continuity and boundary rationality, and provides spatial analysis units with physical meaning.
Smart Images
Figure CN122367147A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of geological disaster risk assessment and geographic information technology, and in particular to a geological disaster susceptibility assessment method that integrates spatial zoning and factor optimization. Background Technology
[0002] Geological hazard susceptibility assessment is a technical method that predicts the spatial distribution probability of hazards by analyzing the characteristics of the hazard-prone environment. The geological hazards covered in this patent mainly refer to slope-type geological hazards (including landslides, mudslides, and debris flows). As the most destructive type of natural hazard in China, the uncertainty in predicting slope-type geological hazards has become a core issue restricting the effectiveness of disaster prevention and mitigation.
[0003] In existing technologies, geological hazard susceptibility assessments generally employ administrative divisions, regular grids, or single-zone methods to spatially divide the study area, and then select a uniform set of environmental factors across the entire study area to construct a predictive model. While this assessment model simplifies the modeling process to some extent, due to significant differences in geological conditions, topographic features, and human engineering activities across different regions, a uniform spatial unit and a uniform combination of characteristic factors cannot accurately reflect the spatial differences in the geological hazard-generating environment.
[0004] Furthermore, the selection of environmental factors in existing technologies typically relies on empirical judgment or existing regional research findings. However, due to the spatial heterogeneity of the geological environment, the combination of disaster-causing factors varies significantly across different regions. The degree of influence of environmental factors on the occurrence of geological disasters in different spatial regions cannot be effectively distinguished, which can easily introduce redundant or weakly correlated factors, thereby affecting the stability and reliability of geological disaster susceptibility assessment results.
[0005] Therefore, how to adaptively construct spatial heterogeneous partitions based on the spatial distribution characteristics of geological disaster events, and dynamically screen environmental factors for collaborative modeling at the partition scale, in order to overcome the shortcomings of existing technologies that ignore spatial heterogeneity and feature differences, has become a technical problem that urgently needs to be solved in this field. Summary of the Invention
[0006] The main objective of this invention is to provide a geological hazard susceptibility assessment method that integrates spatial zoning and factor optimization.
[0007] Another objective of this invention is to propose a geological hazard susceptibility assessment device that integrates spatial zoning and factor optimization.
[0008] The third objective of this invention is to provide a computer device.
[0009] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.
[0010] To achieve the above objectives, a first aspect of the present invention proposes a method for evaluating the susceptibility to geological hazards that integrates spatial zoning and factor optimization, comprising:
[0011] Data on geological hazard points and multi-source environmental factors related to geological hazards in the study area were acquired, and the multi-source environmental factor data were preprocessed to construct a basic dataset for geological hazard susceptibility analysis. Spatial clustering of the geological hazard point data is performed based on the Robust K-means clustering algorithm. The optimal number of clusters K is determined by at least one of the evaluation indicators of elbow rule, silhouette coefficient and Kalinsky-Hallabus index, and the clustering results are obtained. Based on the clustering results and combined with the Thiessen polygon theory, the study area is divided into multiple spatially heterogeneous sub-regions, forming a spatial partitioning system of levels 1 to n, where n>K; For each sub-region, the Fick's Law-Nearest Neighbor model is used to select the optimal subset of feature factors applicable to each sub-region from the base dataset; Based on the particle swarm optimization-random forest model and the optimal feature factor subsets corresponding to each sub-region, the geological hazard susceptibility is predicted for each sub-region. The prediction results of each sub-region are then integrated to obtain the geological hazard susceptibility prediction results for the entire study area.
[0012] In one embodiment of the present invention, the multi-source environmental factor data includes topographic and geomorphological factors, geological structural factors, meteorological and hydrological factors, land cover factors, and human engineering activity factors, wherein: The topographic and geomorphic factors include at least one of elevation, slope, aspect, profile curvature, planar curvature, topographic relief, and roughness. The geological structural factors include at least one of rock mass type, fault density, and distance from fault. The meteorological and hydrological factors include at least one of the following: topographic humidity index (TWI), annual average rainfall, normalized vegetation index (NDVI), distance from river, and river density. The land cover factor includes at least one of land use type and disaster point density; The human engineering activity factors include at least one of the following: distance from road, road density, and population density.
[0013] In one embodiment of the present invention, the preprocessing of the multi-source environmental factor data includes: The multi-source environmental factor data are subjected to coordinate unification, scale matching, and numerical standardization. The Pearson correlation coefficient was used to perform correlation analysis on the multi-source environmental factor data, and strong correlation factors with correlation exceeding a preset threshold were removed. The variance inflation factor is used to detect multicollinearity in the multi-source environmental factor data, and factors with variance inflation factors exceeding a preset threshold are removed; the formula for calculating the variance inflation factor is:
[0014] in, Indicates the first The variance inflation factor of each characteristic factor Indicates the first The coefficient of determination obtained by performing linear regression analysis on one characteristic factor and the remaining characteristic factors.
[0015] In one embodiment of the present invention, the step of using the Fick's Law-Nearest Neighbor model to select the optimal subset of feature factors suitable for each sub-region from the basic dataset includes: Extract the multi-source environmental factor data corresponding to each spatially heterogeneous sub-region; Using disaster and non-disaster samples within the sub-region as training samples, the Fick's law-nearest neighbor model is used to dynamically filter the extracted multi-source environmental factor data to obtain the optimal subset of feature factors suitable for the sub-region.
[0016] In one embodiment of the present invention, the step of dynamically filtering the extracted multi-source environmental factor data using the Fick's Law-Nearest Neighbor model includes: Numerical character feature processing and normalized numerical feature processing are performed on the extracted multi-source environmental factor data; The population initialization and optimization iteration are performed using the Fick's Law-Nearest Neighbor model until the preset termination condition is met. The classification accuracy of the feature subset is tested using the KNN classifier and the ten-fold cross-validation method, and the optimal feature factor subset and related parameters are output.
[0017] In one embodiment of the present invention, the geological hazard susceptibility prediction for each sub-region is performed based on the particle swarm optimization-random forest model and the optimal feature factor subset corresponding to each sub-region, and the prediction results of each sub-region are integrated to obtain the geological hazard susceptibility prediction result for the entire study area, including: Based on the optimal feature factor subsets corresponding to each spatially heterogeneous sub-region, a particle swarm-random forest model is constructed. The key parameters of the random forest model are optimized using the particle swarm optimization algorithm, and geological hazard susceptibility prediction is performed in each sub-region to obtain the hazard susceptibility probability results for each sub-region. The probability results of hazard occurrence in each sub-region are spatially stitched and summarized to obtain the geological hazard occurrence prediction results for the entire study area, and a geological hazard occurrence evaluation map is generated.
[0018] In one embodiment of the present invention, the number of the spatially heterogeneous sub-regions is adaptively determined according to the spatial distribution characteristics of geological hazard points in the study area, and the distribution characteristics and environmental factors of geological hazard points within the same sub-region are relatively consistent, while there are significant differences in the hazard-inducing environment and the main hazard-causing factors between different sub-regions. The spatial distribution characteristics include at least one of the following: the degree of clustering of disaster points, the distribution density, and the degree of spatial dispersion.
[0019] To achieve the above objectives, a second aspect of the present invention provides a geological hazard susceptibility assessment device that integrates spatial zoning and factor optimization, comprising: The data acquisition and preprocessing module is used to acquire geological hazard point data and multi-source environmental factor data related to geological hazards in the study area, and to preprocess the multi-source environmental factor data to construct a basic dataset for geological hazard susceptibility analysis. The spatial clustering module is used to perform spatial clustering of the geological hazard point data based on the Robust K-means clustering algorithm. It determines the optimal number of clusters K by using at least one evaluation index among the elbow rule, silhouette coefficient and Kalinsky-Hallabus index, and obtains the clustering results. The partitioning construction module is used to divide the study area into multiple spatially heterogeneous sub-regions based on the clustering results and in combination with the Thiessen polygon theory, forming a spatial partitioning system of level 1 to n, where n>K; The feature filtering module is used to filter the optimal subset of feature factors suitable for each sub-region from the basic dataset using the Fick's Law-Nearest Neighbor model. The susceptibility prediction module is used to predict the susceptibility of geological hazards for each sub-region based on the particle swarm-random forest model and the optimal feature factor subset corresponding to each sub-region. The prediction results of each sub-region are then integrated to obtain the geological hazard susceptibility prediction results for the entire study area.
[0020] To achieve the above objectives, a third aspect of this application provides a computer device comprising a processor and a memory; wherein the processor runs a program corresponding to the executable program code stored in the memory, for implementing a geological hazard susceptibility assessment method integrating spatial partitioning and factor optimization as described in the first aspect embodiment.
[0021] To achieve the above objectives, the fourth aspect of this application proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a geological hazard susceptibility evaluation method that integrates spatial partitioning and factor optimization as described in the first aspect embodiment.
[0022] The embodiments of the present invention have the following beneficial effects: This invention constructs spatially heterogeneous partitions driven by the spatial distribution characteristics of geological disaster events. This allows the division of spatial units to be dominated by disaster-inducing characteristics, rather than relying on administrative divisions or regular grids. This preserves the nonlinear characteristics of the spatial distribution of geological disasters while ensuring the spatial continuity and reasonable boundaries of each sub-region. The resulting predictive unit system provides spatially meaningful analytical units for subsequent geological disaster susceptibility modeling, effectively overcoming the model bias caused by subjective partitioning in traditional methods.
[0023] This invention introduces a dynamic evolution and collaborative modeling mechanism for characteristic factors at the scale of spatially heterogeneous sub-regions. This enables the selection of environmental factors and the model construction process to be adaptively adjusted for the disaster incubation conditions of different sub-regions, effectively reducing the interference of redundant factors and weakly correlated factors, reducing the impact of spatial heterogeneity on the evaluation results, and thus improving the stability and reliability of the geological disaster susceptibility evaluation results. Attached Figure Description
[0024] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart of a geological hazard susceptibility assessment method that integrates spatial zoning and factor optimization is provided in an embodiment of the present invention; Figure 2 This is a detailed flowchart of a geological hazard susceptibility assessment method that integrates spatial zoning and factor optimization, provided by an embodiment of the present invention. Figure 3 This is a graph showing the variation trend of the silhouette coefficient and CH index under different cluster numbers K in an embodiment of the present invention. Figure 4 The following are analysis charts of SSE, CH index and ΔSSE under different cluster numbers K provided in the embodiments of the present invention; Figure 5 A schematic diagram of the 1-6 level regional division of County A based on Thiessen polygons provided in this embodiment of the invention; Figure 6 This is a diagram showing the optimal feature factor subsets and their feature importance percentages for each sub-region of County A, provided in an embodiment of the present invention. Figure 7 This invention provides a geological hazard susceptibility evaluation map that integrates zoning and feature factor optimization in an embodiment of the invention. Figure 8 This is a structural diagram of a geological hazard susceptibility evaluation device that integrates spatial zoning and factor optimization, provided in an embodiment of the present invention. Detailed Implementation
[0025] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0026] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0027] The following describes, with reference to the accompanying drawings, a geological hazard susceptibility evaluation method that integrates spatial zoning and factor optimization according to an embodiment of the present invention.
[0028] Example 1 This embodiment provides a geological hazard susceptibility assessment method that integrates spatial zoning and factor optimization, and uses County A, which is prone to slope disasters, as the study area for method verification. County A has a complex geological tectonic background and is affected by frequent human engineering activities (mining, road construction, etc.). Geological hazards are mainly landslides, collapses, and debris flows, which are characterized by strong suddenness, short disaster time, and large local damage.
[0029] To compare and analyze the effects of spatial partitioning strategies and feature factor optimization mechanisms, this embodiment further constructs multiple prediction models for comparative verification. The prediction models include: Partitioned feature selection model (PFS): This model performs feature factor optimization on each spatially heterogeneous sub-partition under a spatial partitioning system of 1 to 6 levels, which is the complete technical solution proposed in Embodiment 1 of this invention; Partitioned uniform-feature model (PUF): This model also uses a 1-6 level spatial partitioning system, but uses a uniform combination of feature factors for modeling within each sub-partition, without performing feature factor optimization at the partition scale. Global feature selection model (GFS): This model performs feature factor selection on the entire study area without spatial partitioning, that is, it only uses Fick's law-nearest neighbor model to select feature factors at the global scale. Global uniform-feature model (GUF): This model serves as a control scheme, using a fixed set of feature factors for modeling and prediction without spatial partitioning or feature factor optimization.
[0030] The four prediction models mentioned above differ in their handling of spatial heterogeneity and their feature factor selection mechanisms. The zonal feature selection model and the zonal identical feature model characterize the spatial differences in the disaster-generating environment through spatial zoning, while the global feature selection model and the global identical feature model do not perform spatial zoning. The zonal feature selection model and the global feature selection model introduce a feature factor optimization mechanism based on Fick's law-nearest neighbor model, while the zonal identical feature model and the global identical feature model use a fixed set of feature factors for prediction. Through the construction and application of these various prediction models, a comprehensive analysis of the synergistic and independent effects of spatial heterogeneity zoning strategies and feature factor optimization mechanisms in geological disaster susceptibility assessment is achieved.
[0031] like Figure 1 and Figure 2 As shown, the method includes the following steps: S1. Obtain geological hazard point data and multi-source environmental factor data related to geological hazards in the study area, and preprocess the multi-source environmental factor data to construct a basic dataset for geological hazard susceptibility analysis.
[0032] In this embodiment, geological hazard point data and multi-source environmental factor data related to geological hazards in County A are acquired. The geological hazard point data includes the spatial location and attribute information of slope-type geological hazards such as historical landslides, collapses, and debris flows. Based on the location information of hazard points in County A, non-hazard points are selected in areas more than 1000 meters away from the hazard points through a GIS platform.
[0033] It should be noted that multi-source environmental factor data includes topographic and geomorphological factors, geological structure factors, meteorological and hydrological factors, land cover factors, and human engineering activity factors. Topographic and geomorphological factors include at least one of elevation, slope, aspect, profile curvature, planar curvature, topographic relief, and roughness. Geological structure factors include at least one of rock mass type, fault density, and distance from fault. Meteorological and hydrological factors include at least one of topographic humidity index (TWI), average annual rainfall, normalized difference vegetation index (NDVI), distance from river, and river density. Land cover factors include at least one of land use type and disaster point density. Human engineering activity factors include at least one of distance from road, road density, and population density.
[0034] Next, the acquired multi-source environmental factor data of County A were processed by coordinate unification, spatial resolution and spatial range matching, and continuous factors were normalized and discrete factors were classified and coded to construct a basic dataset for geological disaster susceptibility analysis.
[0035] Furthermore, the correlation between various environmental factors was analyzed using the Pearson correlation coefficient, and redundant factors with excessively strong correlations were eliminated. Generally, a correlation coefficient absolute value exceeding 0.7 indicates a strong correlation between the two factors, and one of the factors should be eliminated. Simultaneously, the variance inflation factor (VIF) was used to detect multicollinearity, eliminating factors with VIF values higher than the threshold of 10. The formula for calculating the variance inflation factor is:
[0036] in, Indicates the first The variance inflation factor of each characteristic factor Indicates the first The coefficient of determination obtained by performing linear regression analysis on one characteristic factor and the remaining characteristic factors.
[0037] S2. Based on the Robust K-means clustering algorithm, spatial clustering is performed on the geological disaster point data. The optimal number of clusters K is determined by at least one of the evaluation indicators of elbow rule, silhouette coefficient and Kalinsky-Hallabus index, and the clustering results are obtained.
[0038] Specifically, after constructing the basic dataset, spatial clustering analysis is performed on the geological hazard points based on their spatial distribution characteristics. Specifically, the Robust K-means clustering algorithm is used to cluster the geological hazard points. The clustering results are comprehensively evaluated using the elbow rule, silhouette coefficient, and Kalinsky-Hallabus index (CH index) to determine the optimal number of clusters K (n>K). The silhouette coefficient is calculated using the following formula:
[0039] In the formula, Indicates sample The average distance to other samples in the same cluster. Indicates sample The minimum average distance to all samples in all other clusters; The closer the value is to 1, the better the clustering effect.
[0040] The formula for calculating the Kalinsky-Harabas Index (CH Index) is as follows:
[0041] in, For the number of clusters, The total number of samples, The sum of squares of inter-cluster deviations The sum of squared deviations within a cluster is expressed as:
[0042] in, For the first Clusters, For the first The center of each cluster, For the first Number of samples in each cluster The CH index serves as the global center for all samples. A larger CH index value indicates greater compactness within clusters, higher separation between clusters, and better clustering performance.
[0043] In this embodiment, the Robust K-means clustering method is used to perform cluster analysis on disaster point samples. The number of clusters K is set to 2~8, and the optimal number of clusters is evaluated by combining the silhouette coefficient, CH index, and elbow rule (SSE). Figure 3 The figure shows the trend of silhouette coefficient and CH index under different cluster numbers K. Figure 4 The figure shows the SSE, CH index, and ΔSSE analysis under different cluster numbers K. When the number of clusters is 3, the rate of SSE decrease slows significantly, entering the "elbow inflection point" range, and the CH index performs exceptionally well, representing the theoretically optimal number of clusters. When the number of clusters is 4, both the silhouette coefficient and the CH index reach their peak values (profile coefficient approximately 0.68, CH index 59.39), indicating optimal intra-cluster compactness and inter-cluster diversity. Based on the spatial distribution characteristics of geological hazards in County A and the verification of zoning rationality, the SSE decrease further narrows when the number of clusters is 4. Further increasing the K value has limited effect on improving cluster compactness; therefore, a cluster number of 4 is ultimately selected as the optimal zoning scheme.
[0044] S3. Based on the clustering results and combined with the Thiessen polygon theory, the study area is divided into multiple spatially heterogeneous sub-regions, forming a spatial partitioning system of level 1 to n, where n>K.
[0045] In this embodiment, based on the clustering results determined in S2 (K=4), the study area is spatially divided using the cluster centers and the Thiessen polygon principle, constructing multiple spatially heterogeneous sub-regions to form a spatial zoning system of levels 1 to n, where n>K. To verify the rationality of K=4, spatial zoning systems of levels 1 to 6 (i.e., n=6) are constructed respectively. By comparing the consistency of the spatial distribution characteristics of geological hazard points and the mechanisms of environmental factors within each sub-region under different numbers of zoning, as well as the differences between sub-regions, the optimal zoning level is determined. Figure 5 The diagram shows the 1-6 level regional division of County A based on Thiessen polygons.
[0046] Specifically, the constructed spatially heterogeneous sub-regions meet the following conditions: the spatial distribution characteristics and environmental factor mechanisms of geological hazard points within the same sub-region are relatively consistent; and there are significant differences in the hazard-inducing environment and the main hazard-causing factors between different sub-regions. The number of sub-regions is adaptively determined based on the spatial distribution characteristics of geological hazard points in County A. The spatial distribution characteristics include the degree of clustering, distribution density, and spatial dispersion of hazard points.
[0047] S4. For each sub-region, the Fick's Law-Nearest Neighbor Model is used to select the optimal subset of feature factors applicable to each sub-region from the basic dataset.
[0048] Specifically, for each spatially heterogeneous sub-region, multi-source environmental factor data corresponding to that sub-region are extracted. Using disaster samples and non-disaster samples within the sub-region as training samples, the Fick's law-nearest neighbor model is used to dynamically filter the extracted multi-source environmental factor data to obtain the optimal subset of feature factors suitable for that sub-region.
[0049] Furthermore, to address the issue of differing geological hazard formation mechanisms within different spatially heterogeneous sub-regions, feature factor optimization was conducted separately for each spatially heterogeneous sub-region. Specifically, corresponding multi-source environmental factor data were extracted within each sub-region, and hazard and non-hazard samples within the sub-region were used as training samples. The Fick's Law-Nearest Neighbor Model (FLA_KNN) was employed to dynamically screen candidate environmental factors.
[0050] It should be noted that the K-Nearest Neighbor (KNN) classifier algorithm is simple and efficient, and is an instance-based nonparametric method. Based on multiple research findings, the KNN classifier has wider applicability compared to more complex classifiers. Therefore, KNN is chosen as the classifier for feature subsets, and the FLA algorithm is combined with the KNN classification algorithm to obtain a new wrapper-style FLA_KNN algorithm, which is applied as an intelligent feature selection mechanism to feature selection problems. The specific steps based on the FLA_KNN algorithm are as follows: (1) Import the extracted multi-source environmental factor data, and perform numerical character feature processing and normalized numerical feature processing on the data; (2) Generate an initial population and use Fick's law FLA algorithm for optimization; (3) Determine whether the preset termination condition is met. If it is met, output the optimal feature subset; otherwise, return to the optimization step. (4) During the optimization process, the KNN classifier and the ten-fold cross-validation method are used to test the classification accuracy of the feature subset; (5) Finally, output the optimal feature subset and the required relevant parameters.
[0051] During feature selection, a feature subset search space is constructed, using classification accuracy as the fitness function. A neighborhood search and perturbation update mechanism are used to continuously optimize feature combinations until a preset termination condition is met, outputting the optimal feature factor subset suitable for the heterogeneous sub-partitions of this space. A neighborhood search strategy generates a neighborhood solution set near the current optimal solution, and a perturbation update mechanism is introduced. The neighborhood search strategy includes, but is not limited to, single-point flipping or feature swapping operations. When the population is detected to be trapped in a local optimum, the perturbation update mechanism randomly resets some feature dimensions to escape the local extremum.
[0052] Specifically, the termination conditions include at least one of the following: reaching the maximum number of iterations, the variation of the optimal fitness over several consecutive generations being less than the convergence threshold, or the classification accuracy reaching a preset target value.
[0053] Furthermore, the K-nearest neighbor classifier is used as the evaluation model, and the classification accuracy of the current feature subset is calculated using K-fold cross-validation. Specifically, the dataset is divided into K mutually exclusive subsets, one subset is selected as the validation set, and the rest are used as the training set. The average value of the K validation results is taken as the fitness value of the feature subset.
[0054] Finally, the optimal feature subset and its corresponding model configuration parameters are output. These model configuration parameters include the optimal fitness value, the number of convergence iterations, and the feature factor identifiers required to construct the spatially heterogeneous subpartition.
[0055] like Figure 6 As shown, this table shows the subsets of optimal feature factors and their proportions of importance for each sub-region of County A. (a) shows the specific composition of the subsets of optimal feature factors for each sub-region, and (b) shows the distribution of the proportions of feature factors importance for each sub-region.
[0056] S5. Based on the particle swarm-random forest model and the optimal feature factor subset corresponding to each sub-region, the geological hazard susceptibility prediction is carried out for each sub-region, and the prediction results of each sub-region are integrated to obtain the geological hazard susceptibility prediction results for the entire study area.
[0057] After obtaining the optimal combination of feature factors for each spatially heterogeneous sub-region, a geological hazard susceptibility prediction model is constructed within each sub-region. Based on the optimal subset of feature factors for each spatially heterogeneous sub-region, a particle swarm optimization-random forest model is constructed. The key parameters of the random forest model are optimized using the particle swarm optimization algorithm, and geological hazard susceptibility prediction is performed in each sub-region to obtain the hazard susceptibility probability results for each sub-region.
[0058] Specifically, based on the random forest model of particle swarm optimization, the key parameters of the random forest model are optimized, and the optimal feature factor subset corresponding to each sub-partition is used as the input variable, and the geological disaster occurrence state is used as the output variable to predict the geological disaster susceptibility of each sub-partition, so as to obtain the disaster susceptibility probability results of each spatially heterogeneous sub-partition.
[0059] The hazard susceptibility probability results of each sub-region are spatially stitched and summarized to form the overall geological hazard susceptibility assessment result for County A. The geological hazard susceptibility probability results obtained within each spatially heterogeneous sub-region are also spatially stitched and summarized to form the overall geological hazard susceptibility assessment result for the study area.
[0060] Furthermore, the study area was divided into levels based on the probability values of susceptibility. To facilitate comparison and evaluation results, the susceptibility was divided into five levels using the equal interval method: low (0-0.2), low-medium (0.2-0.4), medium (0.4-0.6), medium-high (0.6-0.8), and high (0.8-1). Finally, the rationality of the susceptibility mapping and the impact of the zoning feature selection on the evaluation results were analyzed to determine the effectiveness. A geological hazard susceptibility zoning map was generated, as shown below. Figure 7 As shown, this is a geological hazard susceptibility assessment map that integrates zoning and feature factor optimization. (a) shows the results of the GUF model, and (b) to (f) show the results of the PUF2-PUF6 level models, which are used for geological hazard risk analysis and disaster prevention and mitigation decision support. Specific statistical results are shown in Table 1.
[0061] Table 1 Statistical Analysis of Landslide Susceptibility Zones
[0062] Example 2 This invention also provides a geological hazard susceptibility assessment device that integrates spatial zoning and factor optimization, such as... Figure 8 As shown, the device 10 includes: The data acquisition and preprocessing module 100 is used to acquire geological hazard point data and multi-source environmental factor data related to geological hazards in the study area, and to preprocess the multi-source environmental factor data to construct a basic dataset for geological hazard susceptibility analysis. The spatial clustering module 200 is used to perform spatial clustering on the geological hazard point data based on the Robust K-means clustering algorithm, determine the optimal number of clusters K through at least one evaluation index among the elbow rule, silhouette coefficient and Kalinsky-Hallabus index, and obtain the clustering results; The partitioning construction module 300 is used to divide the study area into multiple spatially heterogeneous sub-regions based on the clustering results and in combination with the Thiessen polygon theory, forming a spatial partitioning system of level 1 to n, where n>K; The feature filtering module 400 is used to filter the optimal subset of feature factors suitable for each sub-region from the basic dataset using the Fick's Law-Nearest Neighbor model. The susceptibility prediction module 500 is used to predict the susceptibility of geological disasters for each sub-region based on the particle swarm-random forest model and the optimal feature factor subset corresponding to each sub-region, and integrate the prediction results of each sub-region to obtain the geological disaster susceptibility prediction results for the entire study area.
[0063] Example 3 To implement the methods of the above embodiments, the present invention also provides a computer device, which includes a memory and a processor; wherein the processor runs a program corresponding to the executable program code by reading executable program code stored in the memory, so as to implement the various steps of the methods described above.
[0064] Example 4 To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.
[0065] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
[0066] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0067] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.
Claims
1. A geological hazard susceptibility assessment method integrating spatial zoning and factor optimization, characterized in that, Includes the following steps: Data on geological hazard points and multi-source environmental factors related to geological hazards in the study area were acquired, and the multi-source environmental factor data were preprocessed to construct a basic dataset for geological hazard susceptibility analysis. Spatial clustering of the geological hazard point data is performed based on the Robust K-means clustering algorithm. The optimal number of clusters K is determined by at least one of the evaluation indicators of elbow rule, silhouette coefficient and Kalinsky-Hallabus index, and the clustering results are obtained. Based on the clustering results and combined with the Thiessen polygon theory, the study area is divided into multiple spatially heterogeneous sub-regions, forming a spatial partitioning system of levels 1 to n, where n>K; For each sub-region, the Fick's Law-Nearest Neighbor model is used to select the optimal subset of feature factors applicable to each sub-region from the base dataset; Based on the particle swarm optimization-random forest model and the optimal feature factor subsets corresponding to each sub-region, the geological hazard susceptibility is predicted for each sub-region. The prediction results of each sub-region are then integrated to obtain the geological hazard susceptibility prediction results for the entire study area.
2. The method according to claim 1, characterized in that, The multi-source environmental factor data includes topographic and geomorphological factors, geological structure factors, meteorological and hydrological factors, land cover factors, and human engineering activity factors, among which: The topographic and geomorphic factors include at least one of elevation, slope, aspect, profile curvature, planar curvature, topographic relief, and roughness. The geological structural factors include at least one of rock mass type, fault density, and distance from fault. The meteorological and hydrological factors include at least one of the following: topographic humidity index (TWI), annual average rainfall, normalized vegetation index (NDVI), distance from river, and river density. The land cover factor includes at least one of land use type and disaster point density; The human engineering activity factors include at least one of the following: distance from road, road density, and population density.
3. The method according to claim 1, characterized in that, The preprocessing of the multi-source environmental factor data includes: The multi-source environmental factor data are subjected to coordinate unification, scale matching, and numerical standardization. The Pearson correlation coefficient was used to perform correlation analysis on the multi-source environmental factor data, and strong correlation factors with correlation exceeding a preset threshold were removed. The variance inflation factor is used to detect multicollinearity in the multi-source environmental factor data, and factors with variance inflation factors exceeding a preset threshold are removed; the formula for calculating the variance inflation factor is: in, Indicates the first The variance inflation factor of each characteristic factor Indicates the first The coefficient of determination obtained by performing linear regression analysis on one characteristic factor and the remaining characteristic factors.
4. The method according to claim 1, characterized in that, The Fick's Law-Nearest Neighbor model is used to select the optimal subset of feature factors suitable for each sub-region from the basic dataset, including: Extract the multi-source environmental factor data corresponding to each spatially heterogeneous sub-region; Using disaster and non-disaster samples within the sub-region as training samples, the Fick's law-nearest neighbor model is used to dynamically filter the extracted multi-source environmental factor data to obtain the optimal subset of feature factors suitable for the sub-region.
5. The method according to claim 4, characterized in that, The method of dynamically filtering the extracted multi-source environmental factor data using the Fick's Law-Nearest Neighbor model includes: Numerical character feature processing and normalized numerical feature processing are performed on the extracted multi-source environmental factor data; The population initialization and optimization iteration are performed using the Fick's Law-Nearest Neighbor model until the preset termination condition is met. The classification accuracy of the feature subset is tested using the KNN classifier and the ten-fold cross-validation method, and the optimal feature factor subset and related parameters are output.
6. The method according to claim 1, characterized in that, The method, based on the particle swarm optimization-random forest model and the optimal feature factor subset corresponding to each sub-region, predicts the geological hazard susceptibility of each sub-region separately, and integrates the prediction results of each sub-region to obtain the geological hazard susceptibility prediction results for the entire study area, including: Based on the optimal feature factor subsets corresponding to each spatially heterogeneous sub-region, a particle swarm-random forest model is constructed. The key parameters of the random forest model are optimized using the particle swarm optimization algorithm, and geological hazard susceptibility prediction is performed in each sub-region to obtain the hazard susceptibility probability results for each sub-region. The probability results of hazard occurrence in each sub-region are spatially stitched and summarized to obtain the geological hazard occurrence prediction results for the entire study area, and a geological hazard occurrence evaluation map is generated.
7. The method according to claim 1, characterized in that, The number of spatially heterogeneous sub-regions is adaptively determined based on the spatial distribution characteristics of geological hazard points in the study area, and the distribution characteristics and environmental factors of geological hazard points within the same sub-region are relatively consistent, while there are significant differences in the hazard-inducing environment and the main hazard-causing factors between different sub-regions. The spatial distribution characteristics include at least one of the following: the degree of clustering of disaster points, the distribution density, and the degree of spatial dispersion.
8. A geological hazard susceptibility assessment device integrating spatial zoning and factor optimization, characterized in that, The device includes: The data acquisition and preprocessing module is used to acquire geological hazard point data and multi-source environmental factor data related to geological hazards in the study area, and to preprocess the multi-source environmental factor data to construct a basic dataset for geological hazard susceptibility analysis. The spatial clustering module is used to perform spatial clustering of the geological hazard point data based on the Robust K-means clustering algorithm. It determines the optimal number of clusters K by using at least one evaluation index among the elbow rule, silhouette coefficient and Kalinsky-Hallabus index, and obtains the clustering results. The partitioning construction module is used to divide the study area into multiple spatially heterogeneous sub-regions based on the clustering results and in combination with the Thiessen polygon theory, forming a spatial partitioning system of level 1 to n, where n>K; The feature filtering module is used to filter the optimal subset of feature factors suitable for each sub-region from the basic dataset using the Fick's Law-Nearest Neighbor model. The susceptibility prediction module is used to predict the susceptibility of geological disasters for each sub-region based on the particle swarm-random forest model and the optimal feature factor subset corresponding to each sub-region. The prediction results of each sub-region are then integrated to obtain the geological disaster susceptibility prediction results for the entire study area.
9. A computer device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the geological hazard susceptibility evaluation method that integrates spatial partitioning and factor selection as described in any one of claims 1-7.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements a geological hazard susceptibility evaluation method that integrates spatial partitioning and factor optimization as described in any one of claims 1-7.