Linear traffic engineering risk zone rapid assessment method

Abstract

The invention provides a linear traffic engineering risk zone rapid assessment method, and belongs to the field of geological disaster engineering.The method comprises the steps that data and data of a research area are collected, information amount values of disaster-inducing factors are calculated through an improved information amount model, the disaster-inducing factors with the highest correlation and the lowest contribution amount are eliminated, and the risk zone of the research area is evaluated; the method comprises the following steps: performing preliminary sample data training by adopting an HSVR model, generating a non-landslide point sample added with displacement information, recalculating an information amount value of the non-landslide point sample, inputting data into an SWO and HSVR coupled model for learning and training to obtain a weight coefficient of each disaster-inducing factor, correcting the weight coefficient, and finally realizing data visualization in a GIS (Geographic Information System). According to the method, an improved information amount model, an improved SHAP interpretable model and an SWO-HSVR model are adopted, a high-accuracy geological disaster regional prediction model and linear traffic engineering risk zone rapid evaluation are realized, the influence of geological disasters on linear traffic engineering is reduced, and engineering safety and sustainable development are guaranteed.

Description

Technical Field

[0001] The invention belongs to the field of geological disaster engineering, and in particular relates to a method for quickly assessing the risk zone of a linear transportation engineering project. Background Art

[0002] Landslide susceptibility assessment is the first and most important step in regional landslide risk management and control, providing important decision-making basis for early prevention and early warning of landslide disasters. Therefore, choosing appropriate technical methods to carry out regional landslide susceptibility assessment research is of great significance for landslide disaster warning and prevention. When using the GIS platform ArcGIS for susceptibility assessment, the random point generation tool is usually used to generate non-landslide points. This point may fall in a high-risk area for geological disasters, and there is a certain error;

[0003] Nowadays, detailed geological disaster assessment and early warning system construction are needed along the railway lines, and corresponding geological disaster prevention measures, such as geological exploration, slope support and drainage systems, are taken to ensure the reliability and safety of the railway lines. Therefore, how to quickly and accurately assess the risk zone is of great practical significance. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a method for rapid assessment of risk zones of linear transportation projects, which provides a scientific basis for geological disaster risk management in the planning, design, construction and operation stages of linear transportation projects, and solves the problems of low accuracy of regional prediction models for geological disasters and slow assessment of risk zones of linear transportation projects.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is: a method for rapid assessment of risk zones of linear traffic engineering, comprising the following steps:

[0006] S1. Collect basic geographic data and historical geological disaster statistics of the study area, and reclassify the vectorization and rasterization of each disaster factor after unifying the coordinate system in GIS, establish a multi-level buffer zone for each disaster factor, divide the raster level, and obtain multiple disaster points;

[0007] S2. Calculate the information value of each disaster factor at each disaster point based on the improved information model;

[0008] S3. Calculate the correlation between the disaster-causing factors, and eliminate the disaster-causing factors and their information values ​​whose correlation is greater than a preset threshold;

[0009] S4. Using the improved SHAP interpretable model, the contribution of each hazard factor to the landslide after elimination is calculated, and the hazard factors and their information values ​​whose contribution is less than the preset threshold are eliminated;

[0010] S5. According to the disaster-causing factors that have been eliminated twice, a learning sample library without displacement information is established, the HSVR model is used to train the sample data, and the trained HSVR model is used to obtain the initial susceptibility partition, and non-landslide point samples are generated to construct the final learning sample library;

[0011] S6. According to the final learning sample library, the information value is recalculated using the improved information value model, and the recalculated information value is input into the SWO-HSVR model coupled with the SWO and HSVR models for learning and training to obtain a regional prediction model for geological hazards;

[0012] S7. Obtain the weight coefficient of each disaster-causing factor based on the regional prediction model of geological disasters, correct the weight coefficient, and import it into the GIS system to generate the linear transportation engineering risk zone and complete the rapid assessment of the linear transportation engineering risk zone.

[0013] The beneficial effects of the present invention are as follows: when using GIS technology to evaluate the susceptibility of landslides, the present invention utilizes an improved information volume model, a HSVR model, and a SWO-HSVR model, thereby fully improving the accuracy of susceptibility evaluation, the evaluation speed of linear transportation engineering risk zones, and the accuracy of the regional prediction model for geological disasters, and realizing rapid evaluation of linear transportation engineering risk zones.

[0014] Furthermore, the basic geographic data include factors such as NDVI, average annual rainfall, average annual temperature, DEM, faults, lithology, water system, seismic acceleration, roads, slope, aspect and displacement, and also include deformation values ​​of each grid point.

[0015] Furthermore, the formula of the improved information volume model is as follows:

[0016]

[0017] Among them, I represents the information value, N j represents the number of landslide disasters in each level under the jth hazard factor, N represents the total number of landslide areas, Q j represents the number of grids occupied by the jth disaster factor, Q represents the total number of grids in the analyzed area, a represents the hyperparameter that controls the saturation of the information value when the probability of the disaster factor is high, and b represents the hyperparameter that controls the saturation of the information value when the probability of the disaster factor is low.

[0018] The beneficial effect of the above further scheme is: the present invention introduces hyperparameter a and hyperparameter b into the improved information volume model, and combines hyperparameter a and hyperparameter b to achieve outstanding effect on the information volume value of one or a class of key disaster-causing factors.

[0019] Furthermore, the calculation formula of the improved SHAP interpretable model is as follows:

[0020]

[0021] Among them, κ i represents the contribution of the i-th hazard factor to the landslide, i represents the hazard factor sequence number, P represents the total hazard factor set, S represents the subset of the total hazard factor set except the i-th hazard factor, S∪{i} represents the new set composed of the S set and the i-th hazard factor, k represents the weight of the sample value difference when the S set contains and does not contain the i-th hazard factor, and f x (S) represents the contribution of each hazard factor in the S set, |S| represents the number of factors in the S set, M represents the total number of hazard factors, κ0 represents the initial landslide contribution, x i represents the information value of the i-th disaster factor;

[0022] Furthermore, the S5 comprises the following steps:

[0023] According to the disaster-causing factors that have been eliminated twice, a learning sample library without displacement information is established in the GIS system;

[0024] S502, using the HSVR model to train the sample data in the sample library, using the trained HSVR model to obtain preliminary susceptibility partitions from the study area, and generating non-landslide point samples;

[0025] S503: Place non-landslide point samples and disaster points in the same sample library, and add displacement information to the sample feature column to construct a final learning sample library.

[0026] Furthermore, the calculation formula of the HSVR model is as follows:

[0027]

[0028] Where Y(·) represents the calculation function of the HSVR model, the hyperparameters of the improved information model are a=b=0.5, n represents the number of existing disaster factors, α i represents the contribution of the ith hazard factor, R represents the feasibility hyperparameter, T represents the total number of hazard factors excluding displacement, and x i represents the information value of the i-th disaster factor, y i represents the label of the ith hazard factor represented by the binary extreme value, represents the loss function, K(x i ,x j ) represents the kernel function, x it represents the coupling eigenvalue of disaster factor i and disaster factor t, x jtrepresents the coupling eigenvalue of disaster factor j and disaster factor t;

[0029] When solving the minimum value of function Y(·), the expression of the limiting condition is as follows:

[0030]

[0031] Among them, α i represents the contribution of the ith disaster factor, y i represents the label of the ith hazard factor represented by the two extreme values;

[0032] The loss function The expression is as follows:

[0033]

[0034] Among them, x represents the information value of the disaster factor, and μ represents the loss range hyperparameter.

[0035] The beneficial effect of the above further scheme is as follows: the present invention uses the HSVR model to perform landslide susceptibility zoning before generating landslide points, generates non-landslide points from the landslide susceptibility zoning, avoids collecting landslide areas as negative samples, reduces sample data errors, and improves zoning accuracy.

[0036] Furthermore, the S6 comprises the following steps:

[0037] S601, recalculating the information value using the improved information model according to the final learning sample library;

[0038] S602, according to the recalculated information value, all disaster factors and displacements are calculated using the SWO-HSVR model based on the coupling of the SWO and HSVR models, wherein the loss range hyperparameters, the feasibility hyperparameters and the hyperparameters of the improved information model are iteratively optimized by SWO to obtain the optimal hyperparameters;

[0039] S603, bringing the optimal hyperparameters into the improved information volume model and the HSVR model to calculate the optimal model parameters, and combining the optimal model parameters to obtain a regional prediction model for geological disasters.

[0040] The effective effect of the above further scheme is: when establishing a regional prediction model for geological disasters, the present invention uses an improved information model and a SWO-HSVR model to better fit the nonlinear relationship and the mutual influence between disaster-causing factors, thereby improving the prediction performance and applicability of the model.

[0041] Furthermore, the formula for modifying the weight coefficient is as follows:

[0042]

[0043] Among them, γ i represents the i-th disaster factor, x i represents the information value of the i-th disaster factor, U represents the total number of disaster factors excluding displacement, and γ d represents the displacement weight coefficient, x d The amount of information representing the displacement. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 The figure is a flow chart of the method of the present invention.

[0045] Figure 2 This is the multi-level buffer reclassification of faults in this embodiment.

[0046] Figure 3 This is the multi-level buffering reclassification of the water system in this embodiment.

[0047] Figure 4 This is the slope reclassification in this embodiment.

[0048] Figure 5 This is the susceptibility partition in this embodiment. DETAILED DESCRIPTION

[0049] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0050] Before describing the present invention, the following terms are explained:

[0051] GIS: Geographic Information System;

[0052] Information: Improved information model;

[0053] SHAP: Machine Learning Model Explainability Tool;

[0054] NDVI: Normalized Difference Vegetation Index;

[0055] DEM: Digital Elevation Model;

[0056] SWO: Spider Wasp Optimizer;

[0057] HSVR: Improved support vector regression method.

[0058] Example 1

[0059] like Figure 1 As shown, the present invention provides a method for rapid assessment of linear traffic risk zones, and its implementation method is as follows:

[0060] S1. Collect basic geographic data and historical geological disaster statistics of the study area, and reclassify the vectorization and rasterization of each disaster factor after unifying the coordinate system in GIS, establish a multi-level buffer zone for each disaster factor, divide the raster level, and obtain multiple disaster points.

[0061] In this embodiment, the basic geographic data include: NDVI, average annual rainfall, average annual temperature, DEM, faults, lithology, water system, seismic acceleration, roads, slope, aspect and displacement and other factors, and also include the deformation value of each grid point; the collected basic geographic data are complete in type, wide in range, high in precision and can cover the entire risk zone, the geological disaster data are complete, and include the deformation value of each grid point, and the deformation value of each grid point is obtained through remote sensing data.

[0062] S2. Calculate the information value of each disaster factor at each disaster point based on the improved information model.

[0063] The formula of the improved information volume model is as follows:

[0064]

[0065] Among them, I represents the information value, N j represents the number of landslide disasters in each level under the jth hazard factor, N represents the total number of landslide areas, Q j represents the number of grids occupied by the jth disaster factor, Q represents the total number of grids in the analyzed area, a represents the hyperparameter that controls the saturation of the information value when the probability of the disaster factor is high, and b represents the hyperparameter that controls the saturation of the information value when the probability of the disaster factor is low.

[0066] In this embodiment, the hyperparameters a and b can be solved by SWO, and the combination of the hyperparameters a and b can achieve a prominent effect on one or a class of key disaster-causing factors.

[0067] S3. Calculate the correlation between the disaster-causing factors, and eliminate the disaster-causing factors and their information values ​​whose correlation is greater than a preset threshold.

[0068] In this embodiment, 0.5 is used as the threshold to eliminate disaster factors and their information values ​​whose correlation is greater than the preset threshold. Starting from the correlation between the disaster factors that cause landslides, the mutual influence between the disaster factors required for building the model is reduced, thereby improving the quality of the disaster factors.

[0069] S4. The improved SHAP interpretable model is used to calculate the contribution of each hazard factor to the landslide after elimination, and the hazard factors and their information values ​​whose contribution is less than the preset threshold are eliminated.

[0070] The calculation formula of the improved SHAP interpretable model is as follows:

[0071]

[0072] Among them, κ i represents the contribution of the i-th hazard factor to the landslide, i represents the hazard factor sequence number, P represents the total hazard factor set, S represents the subset of the total hazard factor set except the i-th hazard factor, S∪{i} represents the new set composed of the S set and the i-th hazard factor, k represents the weight of the sample value difference when the S set contains and does not contain the i-th hazard factor, and f x (S) represents the contribution of each hazard factor in the S set, |S| represents the number of factors in the S set, M represents the total number of hazard factors, κ0 represents the initial landslide contribution, x i Represents the information value of the i-th disaster factor.

[0073] In this embodiment, before performing landslide susceptibility zoning, the improved SHAP interpretable model is used to calculate the contribution of each hazard factor to landslide, and the hazard factors with smaller contributions and their information values ​​are eliminated, and the hazard factors with larger contributions are selected for training.

[0074] S5. According to the disaster factors after two eliminations, a learning sample library without displacement information is established, the HSVR model is used to train the sample data, and the trained HSVR model is used to obtain the initial susceptibility partition, and non-landslide point samples are generated to construct the final learning sample library. The implementation method is as follows:

[0075] S501, establishing a learning sample library that does not contain displacement information in the GIS system according to the disaster-causing factors that have been eliminated twice;

[0076] S502, using the HSVR model to train the sample data in the sample library, using the trained HSVR model to obtain preliminary susceptibility partitions from the study area, and generating non-landslide point samples;

[0077] S503: Place non-landslide point samples and disaster points in the same sample library, and add displacement information to the sample feature column to construct a final learning sample library.

[0078] In this embodiment, the calculation formula of the HSVR model is as follows:

[0079]

[0080] Where Y(·) represents the calculation function of the HSVR model, the hyperparameters of the improved information model are a=b=0.5, n represents the number of existing disaster factors, α i represents the contribution of the ith hazard factor, R represents the feasibility hyperparameter, T represents the total number of hazard factors excluding displacement, and x i represents the information value of the i-th disaster factor, y i represents the label of the ith hazard factor represented by the binary extreme value, represents the loss function, K(x i ,x j ) represents the kernel function, x it represents the coupling eigenvalue of disaster factor i and disaster factor t, x jt represents the coupling eigenvalue of disaster factor j and disaster factor t;

[0081] When solving the minimum value of function Y(·), the expression of the limiting condition is as follows:

[0082]

[0083] Among them, α i represents the contribution of the ith disaster factor, y i represents the label of the ith hazard factor represented by the two extreme values;

[0084] The loss function The expression is as follows:

[0085]

[0086] Among them, x represents the information value of the disaster factor, and μ represents the loss range hyperparameter.

[0087] In this embodiment, based on the disaster-causing factors after correlation and contribution are eliminated, a sample library is established in the GIS system. The sample library does not contain displacement information, and the sample data is trained with the HSVR model. The trained HSVR model is used to perform initial susceptibility zoning, and non-landslide samples are randomly generated in low-susceptibility areas.

[0088] In this embodiment, when the function Y(·) takes the minimum value, the contribution of each hazard factor to the landslide is i It is the weight parameter of each disaster factor.

[0089] S6. According to the final learning sample library, the information value is recalculated using the improved information value model, and the recalculated information value is input into the SWO-HSVR model coupled with the SWO and HSVR models for learning and training to obtain a regional prediction model for geological hazards. The implementation method is as follows:

[0090] S601, recalculating the information value using the improved information model according to the final learning sample library;

[0091] S602, according to the recalculated information value, all disaster factors and displacements are calculated using the SWO-HSVR model based on the coupling of the SWO and HSVR models, wherein the loss range hyperparameters, the feasibility hyperparameters and the hyperparameters of the improved information model are iteratively optimized by SWO to obtain the optimal hyperparameters;

[0092] S603, bringing the optimal hyperparameters into the improved information volume model and the HSVR model to calculate the optimal model parameters, and combining the optimal model parameters to obtain a regional prediction model for geological disasters.

[0093] In this embodiment, the formula for recalculating the information value is the same as the formula for susceptibility partitioning in S5.

[0094] In this embodiment, the SWO-HSVR method is used to calculate all disaster factors and displacements, classify the disaster factors, and calculate the classification information value of each factor.

[0095] S7. Obtain the weight coefficient of each disaster-causing factor based on the regional prediction model of geological disasters, correct the weight coefficient, and import it into the GIS system to generate the linear transportation engineering risk zone and complete the rapid assessment of the linear transportation engineering risk zone.

[0096] The formula for the modified weight coefficient is as follows:

[0097]

[0098] Among them, γ i represents the i-th disaster factor, x i represents the information value of the i-th disaster factor, U represents the total number of disaster factors excluding displacement, and γ d represents the displacement weight coefficient, x d The amount of information representing the displacement.

[0099] In this embodiment, basic geographic data and historical geological disaster statistics are first obtained, and the hazard factors are selected based on correlation analysis and calculation of the contribution of each hazard factor to landslides. The HSVR model is used for initial susceptibility zoning, and non-landslide samples are randomly generated in low-susceptibility areas. The hazard factors are classified and the classified information value of each factor is calculated. The information value of each hazard factor is then input into the SWO and HSVR coupled model for learning and training, and the parameters are repeatedly adjusted to obtain a regional prediction model for geological disasters, and a linear transportation engineering risk zone is generated.

[0100] Example 2

[0101] In this example, a national key project under construction in Tibet is selected. The line runs through complex terrain structures such as the Hengduan Mountains and the Qinghai-Tibet Mountains. These geological structures change dramatically, with large-scale faults, folds and earthquake activities. Detailed structural geological surveys and assessments are required to ensure the stability and safety of the railway line. In addition, natural disasters such as landslides, landslides, and mudslides often occur in the Sichuan-Tibet region.

[0102] According to the collected basic geographic data (NDVI, average annual rainfall, average annual temperature, DEM, faults, lithology, water system, earthquake acceleration, roads, slope, slope direction and other disaster factors), it is processed into raster data and vector data in the same coordinate system, and multi-level buffering and reclassification of factors such as water bodies and faults are performed, such as Figure 2 and Figure 3 As shown, Figure 2 Reclassify the fault multi-level buffer. Figure 3 Reclassify the water system multi-level buffer.

[0103] Then all factor layers are rasterized, and data such as temperature and slope are directly reclassified, such as Figure 4 As shown, Figure 4 Reclassify slope.

[0104] Integrate all processed disaster factors and bring them into the improved information volume model to calculate the corresponding information volume value of each grid, and use the GIS median extraction to point function to extract the information volume value of each disaster factor corresponding to the geological disaster point.

[0105] The information value of each disaster factor corresponding to the geological disaster point is calculated through correlation, the correlation between each disaster factor is obtained, and the disaster factors with correlation greater than 0.5 are eliminated.

[0106] Then, the HSVR model is used to generate non-landslide point samples from the area, and displacement information is added to the sample feature column, which is placed in the same sample library as the disaster points.

[0107] By improving the information volume model and recalculating its information volume, the data is input into the SWO-HSVR model for learning and training, the regional prediction model of geological disasters is obtained, and the weight coefficient of each disaster factor is corrected.

[0108] Finally, the grid calculator in GIS is used to process the weight coefficients of the corrected disaster factors, and the weighted sum is used to obtain the risk value of each grid. The natural breakpoint method is then used to divide the grid risk values ​​into five risk levels: very low, low, medium, high, and very high. The linear traffic engineering risk belt is generated to obtain the susceptibility zoning, such as Figure 5 shown.

[0109] In summary, the rapid assessment method for risk zones of linear transportation projects in this scheme is based on the multi-factor conditions of complex linear transportation projects. It is applicable to most geological replication areas, has clear applicable objects, and can be well applied to the rapid assessment of risk zones of various linear transportation projects.

Claims

1. A method for rapid assessment of risk zones in linear transportation projects, characterized in that: The following steps are involved: S1. Collect basic geographic data and historical geological disaster statistics of the study area, and reclassify the vectorization and rasterization of each disaster factor after unifying the coordinate system in GIS, establish a multi-level buffer zone for each disaster factor, divide the raster level, and obtain multiple disaster points; S2. Calculate the information value of each disaster factor at each disaster point based on the improved information model; S3. Calculate the correlation between the disaster-causing factors, and eliminate the disaster-causing factors and their information values ​​whose correlation is greater than a preset threshold; S4. Using the improved SHAP interpretable model, the contribution of each hazard factor to the landslide after elimination is calculated, and the hazard factors and their information values ​​whose contribution is less than the preset threshold are eliminated; S5. According to the disaster-causing factors that have been eliminated twice, a learning sample library without displacement information is established, the HSVR model is used to train the sample data, and the trained HSVR model is used to obtain the initial susceptibility partition, and non-landslide point samples are generated to construct the final learning sample library; S6. According to the final learning sample library, the information value is recalculated using the improved information value model, and the recalculated information value is input into the SWO-HSVR model coupled with the SWO and HSVR models for learning and training to obtain a regional prediction model for geological hazards; S7. Obtain the weight coefficient of each disaster-causing factor based on the regional prediction model of geological disasters, correct the weight coefficient, and import it into the GIS system to generate the linear transportation engineering risk zone and complete the rapid assessment of the linear transportation engineering risk zone.

2. The method for rapid assessment of risk zones of linear traffic engineering according to claim 1 is characterized in that: The formula of the improved information volume model is as follows: Among them, I represents the information value, N j represents the number of landslide disasters in each level under the jth hazard factor, N represents the total number of landslide areas, Q j represents the number of grids occupied by the jth disaster factor, Q represents the total number of grids in the analyzed area, a represents the hyperparameter that controls the saturation of the information value when the probability of the disaster factor is high, and b represents the hyperparameter that controls the saturation of the information value when the probability of the disaster factor is low.

3. The method for rapid assessment of risk zones of linear traffic engineering according to claim 2 is characterized in that: The calculation formula of the improved SHAP interpretable model is as follows: Among them, κ i represents the contribution of the i-th hazard factor to the landslide, i represents the hazard factor sequence number, P represents the total hazard factor set, S represents the subset of the total hazard factor set except the i-th hazard factor, S∪{i} represents the new set composed of the S set and the i-th hazard factor, k represents the weight of the sample value difference when the S set contains and does not contain the i-th hazard factor, and f x (S) represents the contribution of each hazard factor in the S set, |S| represents the number of factors in the S set, M represents the total number of hazard factors, κ0 represents the initial landslide contribution, x i Represents the information value of the i-th disaster factor.

4. The method for rapid assessment of risk zones of linear traffic engineering according to claim 3 is characterized in that: The S5 comprises the following steps: S501, establishing a learning sample library that does not contain displacement information in the GIS system according to the disaster-causing factors that have been eliminated twice; S502, using the HSVR model to train the sample data in the sample library, using the trained HSVR model to obtain preliminary susceptibility partitions from the study area, and generating non-landslide point samples; S503: Place non-landslide point samples and disaster points in the same sample library, and add displacement information to the sample feature column to construct a final learning sample library.

5. The method for rapid assessment of risk zones of linear traffic engineering according to claim 4 is characterized in that: The calculation formula of the HSVR model is as follows: Where Y(·) represents the calculation function of the HSVR model, the hyperparameters of the improved information model are a=b=0.5, n represents the number of existing disaster factors, α i represents the contribution of the ith hazard factor, R represents the feasibility hyperparameter, T represents the total number of hazard factors excluding displacement, and x i represents the information value of the i-th disaster factor, y i represents the label of the ith hazard factor represented by the binary extreme value, represents the loss function, K(x i ,x j ) represents the kernel function, x it represents the coupling eigenvalue of disaster factor i and disaster factor t, x jt represents the coupling eigenvalue of disaster factor j and disaster factor t; When solving the minimum value of function Y(·), the expression of the limiting condition is as follows: Among them, α i represents the contribution of the ith disaster factor, y i represents the label of the ith hazard factor represented by the two extreme values; The loss function The expression is as follows: Among them, x represents the information value of the disaster factor, and μ represents the loss range hyperparameter.

6. The method for rapid assessment of risk zones of linear traffic engineering according to claim 5 is characterized in that: The S6 comprises the following steps: S601, recalculating the information value using the improved information model according to the final learning sample library; S602, according to the recalculated information value, all disaster factors and displacements are calculated using the SWO-HSVR model based on the coupling of the SWO and HSVR models, wherein the loss range hyperparameters, the feasibility hyperparameters and the hyperparameters of the improved information model are iteratively optimized by SWO to obtain the optimal hyperparameters; S603, bringing the optimal hyperparameters into the improved information volume model and the HSVR model to calculate the optimal model parameters, and combining the optimal model parameters to obtain a regional prediction model for geological disasters.

7. The method for rapid assessment of risk zones of linear traffic engineering according to claim 1 is characterized in that: The formula for the modified weight coefficient is as follows: Among them, γ i represents the i-th disaster factor, x i represents the information value of the i-th disaster factor, U represents the total number of disaster factors excluding displacement, and γ d represents the displacement weight coefficient, x d The amount of information representing the displacement.

Images