Landslide disaster risk dynamic assessment method and system based on improvement project progress

Abstract

The invention discloses a landslide disaster risk dynamic assessment method and system based on an improvement project progress, and relates to the technical field of natural disaster treatment. According to the invention, through integration of multi-source heterogeneous data, real-time dynamic assessment and a machine learning model, accurate and dynamic assessment of landslide disaster risks is realized, assessment precision and stability are improved, and timeliness and pertinence of decision making are enhanced; a feature disturbance stability mechanism is introduced into the SPY algorithm, and a pseudo-negative sample screening strategy with higher robustness and discriminability is constructed; an information entropy reduction amount is introduced, an acquisition function is improved, the efficiency of hyper-parameter optimization is improved, and a stacking generalization integration model is effectively trained; the influence of the regulation situation progress in the landslide disaster area is considered, the susceptibility correction coefficient is set, the landslide disaster susceptibility evaluation result is corrected, it is ensured that the landslide disaster susceptibility evaluation result can reflect the actual situation in time, and powerful technical support is provided for prevention and management of landslide disasters.

Description

Technical Field

[0001] The present invention relates to the technical field of natural disaster treatment, and particularly relates to a dynamic landslide disaster risk assessment method and system based on the progress of improvement projects. Background Art

[0002] At present, landslides are common and extremely destructive geological disasters, seriously threatening the safety of residents' lives and property, the stability of infrastructure, and the sustainable development of regions. Existing landslide risk assessment technologies mainly focus on natural environmental factors such as terrain, geology, and meteorology, generally ignoring the key role of the disaster adaptability and resilience of villages in landslide disasters, resulting in limitations in the applicability of assessment results in actual disaster response, and may lead to unreasonable planning of prevention and control budget funds. There is an urgent need to develop a dynamic landslide risk assessment system that integrates natural environmental factors and social improvement resilience factors to improve the applicability of landslide disaster risk assessment results. Summary of the Invention

[0003] The purpose of the present invention is to provide a dynamic landslide disaster risk assessment method and system based on the progress of improvement projects to address the above technical problems.

[0004] To achieve the above invention purpose, the embodiments of the present invention provide the following technical solutions:

[0005] A dynamic landslide disaster risk assessment method based on the progress of improvement projects, which includes:

[0006] Construct a set of slope units and a dataset for comprehensive landslide disaster susceptibility evaluation in the area to be measured, and divide the positive sample set of landslide disaster hidden danger points;

[0007] Use the SPY-FPS pseudo-negative sample screening method to generate a training set of identified pseudo-negative samples, and train a classifier to obtain an identified pseudo-negative sample generator;

[0008] Input the dataset for comprehensive landslide disaster susceptibility evaluation into the identified pseudo-negative sample generator to obtain an identified pseudo-negative sample set;

[0009] Merge the identified pseudo-negative sample set and the positive sample set of landslide disaster hidden danger points and input them into a stacked generalization integration model to obtain the landslide disaster susceptibility evaluation result;

[0010] Obtain the landslide disaster improvement progress data in real time, construct a susceptibility correction coefficient, and correct the landslide disaster susceptibility evaluation result;

[0011] Based on the corrected landslide disaster susceptibility evaluation result, conduct risk division on the area to be measured.

[0012] Further, constructing a set of slope units and a dataset for comprehensive landslide disaster susceptibility evaluation in the area to be measured, and dividing the positive sample set of landslide disaster hidden danger points, includes:

[0013] The slope unit extraction method using forward and reverse hydrological partitioning and topographic curvature constraints is used to process the area to be measured, and a slope unit set is obtained;

[0014] Collect the initial susceptibility assessment data corresponding to the slope unit set, and connect the initial susceptibility assessment data and the slope unit set through a spatial join tool to obtain an initial dataset for the overall susceptibility assessment of landslide disasters; the initial susceptibility assessment data includes topographic data, geological data, meteorological data, and human activity data for each slope unit;

[0015] Use the Spearman rank correlation coefficient to process the initial dataset for the overall susceptibility assessment of landslide disasters to obtain a positive sample set of potential landslide hazard points and a dataset for the overall susceptibility assessment of landslide disasters.

[0016] Furthermore, the process of obtaining the slope unit set is as follows:

[0017] Collect high-precision DEM data of the area to be measured and input it into a geographic information system for depression filling processing to construct a high-precision digital elevation model;

[0018] Calculate the flow direction and flow accumulation of each grid in the high-precision digital elevation model;

[0019] Based on the flow accumulation, screen each grid to obtain the screened grids;

[0020] According to the flow direction of each screened grid, use the Watershed tool to divide the catchment area and generate a forward division result;

[0021] Perform inversion processing on the high-precision DEM data to obtain high-precision inverted DEM data; process the high-precision inverted DEM data to generate a reverse division result;

[0022] Use the Union tool to merge the boundaries and unify the space of the forward division result and the forward division result to generate a slope boundary framework;

[0023] Calculate the topographic profile curvature and planar curvature of the slope boundary framework and perform regional elimination to obtain the slope boundary framework after elimination;

[0024] Use the polygon clipping tool to divide the area in the slope boundary framework after elimination to form a slope unit set.

[0025] Furthermore, the training process of the classifier includes:

[0026] Obtain a training dataset for the overall susceptibility assessment of landslide disasters, that is, obtain a positive sample training set and an unlabeled sample training set;

[0027] Randomly select several positive samples from the positive sample training set and use them as the embedded reference positive sample training set; use the unselected positive samples in the positive sample training set as the new positive sample training set;

[0028] Merge the embedded reference positive sample training set and the unlabeled sample training set to obtain the recognition pseudo-negative sample training set;

[0029] Input the recognition pseudo-negative sample training set and the new positive sample training set into the classifier for training respectively to obtain the initially trained classifier;

[0030] Obtain the classification probability of the embedded reference positive sample training set through the initially trained classifier and set the confidence threshold;

[0031] Based on the confidence threshold and the unlabeled sample training set, use the feature perturbation stability mechanism to generate a reliable negative sample training set and input it into the initially trained classifier for secondary training to obtain the trained classifier, that is, the recognition pseudo-negative sample generator.

[0032] Furthermore, the stacking generalization ensemble model includes a cascaded base model and a meta-model; the base model includes a decision tree, a random forest, a support vector machine, and a gradient boosting tree that are processed in parallel; the meta-model is a meta-learner;

[0033] Train the stacking generalization ensemble model using the Bayesian hyperparameter optimization method guided by information entropy.

[0034] Furthermore, the Bayesian hyperparameter optimization method guided by information entropy includes:

[0035] Obtain the recognition pseudo-negative sample training set and the landslide hazard point positive sample training set and input them into the stacking generalization ensemble model;

[0036] Define the hyperparameter combination of the stacking generalization ensemble model, construct the corresponding hyperparameter space; set the objective function;

[0037] Construct a learnable surrogate model and set the information entropy acquisition function; determine the optimal hyperparameter combination through the learnable surrogate model and the information entropy acquisition function;

[0038] Apply the optimal hyperparameter combination to the stacking generalization ensemble model to complete the training of the stacking generalization ensemble model.

[0039] Furthermore, the process of determining the optimal hyperparameter combination is:

[0040] Randomly select at least one hyperparameter combination and calculate the corresponding objective function value;

[0041] Based on the selected hyperparameter combination and the objective function value, use Gaussian process regression to construct a learnable surrogate model;

[0042] Set the information entropy acquisition function through a learnable proxy model; calculate the information entropy acquisition function values for all hyperparameter combinations; select the hyperparameter combinations whose information entropy acquisition function values meet the screening conditions as the evaluation points for the next iteration;

[0043] Apply the evaluation points to the stacked generalization integration model, calculate the corresponding objective function values, and update the learnable proxy model;

[0044] Repeat the process of selecting hyperparameter combinations, calculating the corresponding objective function values, and updating the learnable proxy model until the current iteration number reaches the maximum iteration number or the objective function value is less than the objective threshold to obtain the optimal hyperparameter combination.

[0045] Furthermore, the formula corresponding to the information entropy acquisition function is:

[0046] ;

[0047] ;

[0048] Where and represent the second original acquisition function and the first original acquisition function respectively, represents the hyperparameter combination, represents the information entropy reduction, represents the optimal loss value, and represent the variance function and the mean function respectively, represents the intermediate parameter, and represent the cumulative distribution function of the standard normal distribution and the probability density function of the standard normal distribution respectively, represents the logarithmic function with a constant base, and represent the variance in the previous iteration and the variance in the current iteration respectively.

[0049] Furthermore, the real-time acquisition of landslide disaster treatment progress data, construction of a susceptibility correction coefficient, and correction of the landslide disaster susceptibility evaluation results include:

[0050] Obtain the village landslide treatment points and treatment progress in the area to be measured;

[0051] Input the village landslide treatment points into the geographic information system for attribute connection to construct a landslide disaster treatment dataset;

[0052] Perform a spatial connection on the landslide disaster treatment dataset and the landslide disaster susceptibility evaluation results to obtain a landslide-prone area treatment set;

[0053] According to the progress of the renovation of each slope renovation unit in the landslide-prone area, set the percentage of the engineering renovation progress as the corresponding susceptibility correction coefficient;

[0054] Based on the susceptibility correction coefficient, update the landslide hazard susceptibility evaluation results corresponding to each slope renovation unit in the landslide-prone area renovation concentration.

[0055] A dynamic landslide hazard risk assessment system based on the progress of renovation projects includes:

[0056] A slope unit construction module, used to construct a set of slope units in the area to be measured and a landslide hazard global susceptibility evaluation data set, and divide the positive sample set of landslide hazard hidden danger points;

[0057] An identification pseudo-negative sample generation training module, used to generate an identification pseudo-negative sample training set by using the SPY-FPS pseudo-negative sample screening method, and train a classifier to obtain an identification pseudo-negative sample generator;

[0058] An identification pseudo-negative sample acquisition module, used to input the landslide hazard global susceptibility evaluation data set into the identification pseudo-negative sample generator to obtain an identification pseudo-negative sample set;

[0059] A landslide hazard susceptibility evaluation module, including a stacked generalization integration model training module and a landslide hazard susceptibility evaluation prediction module;

[0060] A stacked generalization integration model training module, used to input the landslide hazard susceptibility evaluation training data set into the stacked generalization integration model, and train it by using the Bayesian hyperparameter optimization method guided by information entropy;

[0061] A landslide hazard susceptibility evaluation prediction module, used to input the combined identification pseudo-negative sample set and the positive sample set of landslide hazard hidden danger points into the stacked generalization integration model to obtain the landslide hazard susceptibility evaluation result;

[0062] A landslide hazard susceptibility evaluation result correction module, used to obtain the landslide hazard renovation progress data in real time, construct a susceptibility correction coefficient and correct the landslide hazard susceptibility evaluation result;

[0063] A landslide risk division module, used to divide the risk of the area to be measured based on the corrected landslide hazard susceptibility evaluation result.

[0064] The beneficial effects of the present invention are:

[0065] The present invention realizes the accurate and dynamic assessment of landslide disaster risks by integrating multi-source heterogeneous data, real-time dynamic assessment, and machine learning models, improves the assessment accuracy and stability, enhances the timeliness and pertinence of decision-making; introduces the "feature perturbation stability" mechanism into the SPY (Spy Sample-Based Semi-Supervised Learning Algorithm) method to construct a more robust and discriminative pseudo-negative sample screening strategy, enabling effective identification of reliable negative samples in the context of extremely imbalanced positive and negative samples; introduces the information entropy reduction amount to improve the acquisition function, enhances the efficiency of hyperparameter optimization, and effectively trains the stacked generalization ensemble model; considers the impact of the progress of rectification projects in landslide disaster areas, sets the susceptibility correction coefficient, and corrects the landslide disaster susceptibility evaluation results to ensure that the landslide disaster susceptibility evaluation results can timely reflect the actual situation, providing strong technical support for the prevention and management of landslide disasters. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for the embodiments. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can be obtained based on these drawings.

[0067] Figure 1 It is the flowchart of the method in the embodiment of the present invention;

[0068] Figure 2 It is the system flowchart in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0069] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. The components of the embodiments of the present invention usually described and illustrated in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed present invention, but only represents the selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0070] Please refer to Figure 1 , a method for dynamically assessing the risk of landslide disasters based on the progress of rectification projects provided in this embodiment, which includes:

[0071] S1. Construct the slope unit set of the area to be tested and the global landslide susceptibility assessment data set, and divide the positive sample set of landslide hazard risk points;

[0072] The S1 includes:

[0073] S1-1. The slope unit extraction method based on positive and negative hydrological division and terrain curvature constraint is used to process the area to be tested and obtain a slope unit set.

[0074] The slope unit extraction method with forward and reverse hydrological division and terrain curvature constraint is implemented in a geographic information system, so that S1-1 includes:

[0075] S1-1-1. Collect high-precision DEM data of the area to be measured, input it into the geographic information system for depression filling, and construct a high-precision digital elevation model;

[0076] S1-1-2. Calculate the flow direction and flow accumulation of each grid in the high-precision digital elevation model;

[0077] S1-1-3, determine whether the flow accumulation of each grid exceeds the flow threshold; if so, the grid is regarded as the main water system and retained; otherwise, the grid is discarded; and the screened grid is obtained;

[0078] S1-1-4. Based on each main water system and according to the flow direction of each screened grid, use the watershed tool to divide the catchment area and generate the positive preliminary slope boundary, i.e. the positive division result;

[0079] S1-1-5, invert the high-precision DEM data to obtain high-precision inverted DEM data; use the same method as S1-1-2 to S1-1-4 to process the high-precision inverted DEM data to generate a reverse preliminary slope boundary, that is, a reverse division result, which can enhance the accuracy of landform structure recognition;

[0080] S1-1-6. Use the Union tool to merge the boundaries and unify the spaces of the forward division results and the forward division results to generate a slope boundary framework;

[0081] S1-1-7. Calculate the terrain profile curvature and plane curvature of the slope boundary frame; eliminate the area boundaries where the terrain profile curvature and plane curvature do not exceed the curvature threshold condition to obtain the slope boundary frame after elimination; the curvature threshold condition includes the terrain profile curvature threshold and the plane curvature threshold.

[0082] S1-1-8. Use the planar cutting (Split) tool to divide the area in the slope boundary framework after rejection to obtain the division result; take each division result as a slope unit to form a slope unit set. The expression of the slope unit set is . Among them, , , respectively represent the first slope unit, the second slope unit and the last slope unit, represents the slope unit set, represents the number of slope units.

[0083] S1-2. Collect the initial susceptibility assessment data corresponding to the slope unit set, and use the Spatial Join tool to connect the initial susceptibility assessment data and the corresponding slope units to obtain the initial dataset for the overall susceptibility evaluation of landslide disasters.

[0084] The initial susceptibility assessment data includes the topographic data, geological data, meteorological data and human activity data of each slope unit; the topographic data includes elevation, slope, aspect, curvature and river channel; the geological data includes soil type, rock formation structure, fault distribution and seismic motion parameters; the meteorological data includes the previous cumulative precipitation, precipitation intensity, daily temperature difference, groundwater level; the human activity data includes land use type, road network, underground engineering area.

[0085] The S1-2 includes:

[0086] S1-2-1. Organize the format of the initial susceptibility assessment data, impute missing values, process outliers and encode the data to obtain the susceptibility assessment data;

[0087] S1-2-2. Use the Spatial Join tool to dynamically connect each slope unit in the slope unit set and the susceptibility assessment data to obtain the initial dataset for the overall susceptibility evaluation of landslide disasters . Among them, represents the initial dataset for the overall susceptibility evaluation of landslide disasters, , , respectively represent the first initial data for the overall susceptibility evaluation of landslide disasters, the second initial data for the overall susceptibility evaluation of landslide disasters and the last initial data for the overall susceptibility evaluation of landslide disasters, represents the amount of the initial data for the overall susceptibility evaluation of landslide disasters.

[0088] Any initial data for the overall susceptibility evaluation of landslide disasters includes the susceptibility assessment data of the corresponding slope unit and its pixel raster.

[0089] S1-3. Process the initial dataset of the landslide disaster global susceptibility assessment using the Spearman rank correlation coefficient to obtain the positive sample set of landslide disaster potential points and the dataset for the landslide disaster global susceptibility assessment.

[0090] The above S1-3 includes:

[0091] S1-3-1. Take the slope units containing potential points as the positive samples of landslide disaster potential points to obtain the positive sample set of landslide disaster potential points.

[0092] Call the Spatial Join tool to join the slope unit set and the initial data of the landslide disaster global susceptibility assessment to obtain the joined initial data of the landslide disaster global susceptibility assessment; create the attribute field "Landslide (judgment character)" corresponding to each slope unit in the joined initial data of the landslide disaster global susceptibility assessment and assign judgment attribute information, that is, assign 1 (representing the "positive sample set") to the slope units containing potential points, and assign 0 (representing the "unlabeled sample set") to the units without potential points.

[0093] S1-3-2. According to the formula:

[0094] ;

[0095] Calculate the judgment character of the joined initial data of the landslide disaster global susceptibility assessment . represents the joined initial data of the landslide disaster global susceptibility assessment in the th data. represents the positive sample set of landslide disaster potential points. represents the rd positive sample of landslide disaster potential points.

[0096] S1-3-3. Optimize the initial dataset of the landslide disaster global susceptibility assessment using the Spearman rank correlation coefficient to obtain the dataset for the landslide disaster global susceptibility assessment.

[0097] Take the initial data of the landslide disaster global susceptibility assessment as the susceptibility assessment factors, and use the Spearman rank correlation coefficient analysis method to calculate the correlation coefficients between the susceptibility assessment factors; eliminate the susceptibility assessment factors with collinearity (VIF (correlation coefficient) ≥ 10) in the initial dataset of the landslide disaster global susceptibility assessment to obtain the dataset for the landslide disaster global susceptibility assessment . . . respectively represent the first landslide disaster global susceptibility evaluation data, the second landslide disaster global susceptibility evaluation data, and the last landslide disaster global susceptibility evaluation data, represent the landslide disaster global susceptibility evaluation data set.

[0098] The landslide disaster global susceptibility evaluation data set includes a positive sample set and an unlabeled sample set .

[0099] S2. Use the SPY-FPS pseudo-negative sample screening method to generate a recognition pseudo-negative sample training set, and train a classifier to obtain a recognition pseudo-negative sample generator;

[0100] The training process of the classifier includes:

[0101] S2-1. Obtain a positive sample training set (with a positive label) and an unlabeled sample training set (without a label);

[0102] Obtain a positive sample training set for landslide disaster potential points; demarcate the slope units that intersect with the positive sample training set of landslide disaster potential points as positive samples (with a positive label), and demarcate the slope units that do not contain the positive sample training set of landslide disaster potential points as an unlabeled sample set (without a label);

[0103] S2-2. Randomly select several positive samples from the positive sample training set and use them as an embedded reference positive sample training set ; use the positive samples in the positive sample training set that have not been selected as a new positive sample training set . 20% of the positive sample training set can be randomly selected , that is . Among them, represents the absolute value.

[0104] S2-3. Combine the embedded reference positive sample training set and the unlabeled sample training set to obtain a recognition pseudo-negative sample training set ;

[0105] S2-4. Input the recognition pseudo-negative sample training set and the new positive sample training set into the classifier for training to obtain an initially trained classifier; among them, the classifier can use a CNN classifier or a naive Bayes classifier, and its output is a classification probability. The classification probability includes the probabilities of classifying as positive samples and negative samples.

[0106] S2-5. Obtain the embedded reference positive sample training set through the initially trained classifier The classification probability of the embedded reference positive sample prediction probability set is obtained ; Based on the probability distribution characteristics of the embedded reference positive sample prediction probability set, combined with its skewness and quantile characteristics, analyze it and set the confidence threshold based on the analysis results For example, choose The lower 5% quantile of is taken as the confidence threshold.

[0107] S2-6. Based on confidence threshold , unlabeled sample training set , a reliable negative sample training set is generated by using the feature perturbation stability mechanism, and is input into the initially trained classifier for secondary training to obtain a trained classifier, i.e., an identification pseudo negative sample generator.

[0108] The S2-6 includes:

[0109] S2-6-1. Prediction probability set and confidence threshold based on unlabeled samples , for the unlabeled sample training set Screening to obtain a preliminary pseudo-negative sample training set ; Keep the prediction probability of unlabeled samples not less than the confidence threshold Unlabeled samples with values less than the confidence threshold are removed Unlabeled samples are obtained to obtain the initial pseudo negative sample training set .

[0110] S2-6-2. Introduce feature perturbation stability mechanism to the initial pseudo-negative sample training set Each sample in is feature perturbed to generate R perturbed samples, that is, , construct a more robust and discriminative pseudo-negative sample screening strategy, which can effectively identify reliable negative samples in data scenarios with extremely unbalanced positive and negative samples. Represents the initial pseudo negative sample training set No. samples, Indicates perturbation samples, represents the perturbation function (characteristic perturbation stability mechanism).

[0111] S2-6-3. Input each perturbation sample into the initial classifier respectively, count the number of times the perturbation sample is judged as a negative label, and calculate the corresponding robustness score; the initial classifier can use a naive Bayes classifier.

[0112] S2-6-4. Sort the perturbation samples in descending order according to the robustness scores, and select the perturbation samples corresponding to the top K robustness scores to obtain the final pseudo-negative sample training set. .

[0113] S2-6-5. Combine the final pseudo-negative sample training set and the positive sample training set to obtain a reliable negative sample training set;

[0114] S2-6-6. Input the reliable negative sample training set into the initially trained classifier for secondary training to obtain a trained classifier, that is, an identification pseudo-negative sample generator.

[0115] S3. Input the landslide disaster global susceptibility evaluation data set into the identification pseudo-negative sample generator to obtain an identification pseudo-negative sample set.

[0116] Input the landslide disaster global susceptibility evaluation data set into the identification pseudo-negative sample generator to obtain the corresponding classification probability; select the landslide disaster global susceptibility evaluation data greater than the confidence threshold to obtain an identification pseudo-negative sample set.

[0117] S4. Combine the identification pseudo-negative sample set and the landslide disaster hidden danger point positive sample set and input them into the stacking generalization integration model to obtain the landslide disaster susceptibility evaluation result.

[0118] In this embodiment, a stacking generalization integration model (Stacking integration model) is used for susceptibility evaluation; the stacking generalization integration model includes a cascaded base model and a meta-model; the base model includes a decision tree, a random forest, a support vector machine, and a gradient boosting tree for parallel processing; the meta-model is a meta-learner.

[0119] Use the Bayesian hyperparameter optimization method guided by information entropy to train the stacking generalization integration model; thus, the Bayesian hyperparameter optimization method guided by information entropy includes:

[0120] S4-1. Obtain the identification pseudo-negative sample training set and the landslide disaster hidden danger point positive sample training set and input them into the stacking generalization integration model;

[0121] S4-2. Define the hyperparameter combination of the stacking generalization integration model , construct the corresponding hyperparameter space ; set the initial values of the objective function, the current iteration number, and the maximum iteration number; the formula corresponding to the objective function is:

[0122] ;

[0123] Among them, denotes the objective function, represents the AUC metric of the stacked generalization ensemble model. The objective function aims to maximize the AUC metric during the iterative process.

[0124] Hyperparameter space includes several groups of hyperparameter combinations , and each group of hyperparameter combinations includes the tunable parameters of all base models and the meta-model. The initial value of the current iteration count is 0.

[0125] S4-3. Construct a learnable surrogate model and set the information entropy acquisition function; determine the optimal hyperparameter combination through the learnable surrogate model and the information entropy acquisition function;

[0126] The S4-3 includes:

[0127] S4-3-1. Randomly select at least one hyperparameter combination and calculate the corresponding objective function value;

[0128] S4-3-2. Based on the selected hyperparameter combination and the objective function value, construct a learnable surrogate model using Gaussian process regression ; where, represents the kernel function, represents the selected hyperparameter combination, represents the mean function, represents the Gaussian function;

[0129] S4-3-3. Set the information entropy acquisition function through the learnable surrogate model; calculate the information entropy acquisition function values of all hyperparameter combinations; select the hyperparameter combinations whose information entropy acquisition function values meet the screening conditions as the evaluation points for the next iteration;

[0130] The information entropy acquisition function is the expected improvement acquisition function based on information gain.

[0131] The S4-3-3 includes:

[0132] A1. Take the selected hyperparameter combination and the objective function value as historical data ; where, represents the number of selected hyperparameter combinations, , , respectively represent the first hyperparameter combination, the second hyperparameter combination, and the last hyperparameter combination in the selected hyperparameter combinations, , , respectively represent , , the corresponding objective function values.

[0133] A2. Calculate the similarity between pairs of historical data, the posterior variance and the posterior mean of the current iteration through a learnable proxy model;

[0134] Calculate the similarity between every two historical data using the kernel function in the learnable proxy model, that is ; , respectively represent the th historical data and the th historical data.

[0135] Based on each similarity, calculate the variance and mean of the current iteration through the learnable proxy model.

[0136] A3. Calculate the optimal loss value (optimal objective loss value) of the current iteration, and construct an information entropy acquisition function based on the variance and mean;

[0137] The current iteration of the optimal loss value The corresponding formula is:

[0138] ;

[0139] The first original acquisition function The corresponding formula is:

[0140] ;

[0141] Among them, represents the combination of hyperparameters selected corresponding to the optimal loss value of the current iteration, represents the minimum value function, , respectively represent the variance function and the mean function, represents the intermediate parameter, , respectively represent the cumulative distribution function of the standard normal distribution and the probability density function of the standard normal distribution.

[0142] To improve the efficiency of hyperparameter optimization, introduce the information entropy reduction . The formula corresponding to the information entropy reduction is:

[0143] ;

[0144] Among them, , respectively represent the uncertainty of the hyperparameter space in the previous iteration and the uncertainty of the hyperparameter space in the current iteration.

[0145] , Posterior entropy calculation using Gaussian process regression, thus, The corresponding formula is:

[0146] ;

[0147] ;

[0148] Among them, represents the natural constant, represents the logarithmic function with a constant as the base, represents uncertainty, , represent the variance in the previous iteration and the variance in the current iteration, respectively.

[0149] Thus, the information entropy acquisition function The corresponding formula is:

[0150] ;

[0151] ;

[0152] Among them, represents the second original acquisition function.

[0153] A4. Calculate the information entropy acquisition function values corresponding to all hyperparameter combinations in the hyperparameter space;

[0154] A5. Select the hyperparameter combinations whose information entropy acquisition function values meet the screening conditions as the evaluation points for the next iteration, and merge them into the historical data to complete the update of the historical data; the screening conditions are:

[0155] ;

[0156] Among them, represents the maximum value function.

[0157] S4-3-4. Apply the evaluation points to the stacked generalization ensemble model, calculate the corresponding objective function values, and update the learnable surrogate model; increment the current iteration count by 1, that is ;

[0158] S4-3-5. Repeat S4-3-3 to S4-3-4 until the current iteration count reaches the maximum iteration count or the objective function value is less than the objective threshold, obtain the optimal hyperparameter combination and apply it to the stacked generalization ensemble model to complete the training of the stacked generalization ensemble model. The objective threshold is 0.95. When the objective function value exceeds 0.95, package the stacked generalization ensemble model and process the slope susceptibility evaluation dataset.

[0159] S4-4. Apply the optimal hyperparameter combination to the stacked generalization integration model to complete the training of the stacked generalization integration model.

[0160] S5. Obtain the progress data of landslide disaster treatment in real time, construct a susceptibility correction coefficient, and correct the landslide disaster susceptibility evaluation results.

[0161] The S5 includes:

[0162] S5-1. Obtain the village landslide treatment points and the progress of the treatment situation in the area to be measured, that is, the longitude and latitude of the landslide disaster treatment points and the percentage of the progress of the project carried out . represents longitude, represents latitude.

[0163] S5-2. Input the village landslide treatment points into the geographic information system for attribute connection to construct a landslide disaster treatment dataset . . respectively represent the longitude and latitude of the th village landslide treatment point, represents the progress of the treatment situation of the th village landslide treatment point, represents the total number of village landslide treatment points.

[0164] S5-3. Use the same method as S1-3 to perform spatial connection on the landslide disaster treatment dataset and the landslide disaster susceptibility evaluation results to obtain a treatment set for landslide-prone areas where engineering treatment has been carried out . . . respectively represent the landslide disaster susceptibility evaluation results corresponding to the first slope unit, the second slope unit, and the last slope unit, . . respectively represent the first slope treatment unit, the second slope treatment unit, and the last slope treatment unit in the treatment set for landslide-prone areas where engineering treatment has been carried out.

[0165] S5-4. According to the progress of the treatment situation corresponding to each slope treatment unit in the treatment set for landslide-prone areas, set the percentage of the engineering treatment progress as the corresponding susceptibility correction coefficient; among them, the value range of the susceptibility correction coefficient is [0,1].

[0166] S5-5. Update the landslide disaster susceptibility evaluation results corresponding to each slope treatment unit in the landslide-prone area rectification concentration based on the susceptibility correction coefficient to obtain the corrected landslide disaster susceptibility evaluation results. The corresponding formula is:

[0167] ;

[0168] where, represents the susceptibility correction coefficient corresponding to the th slope treatment unit in the landslide-prone area rectification concentration, , respectively represent the landslide disaster susceptibility evaluation results and the corrected landslide disaster susceptibility evaluation results corresponding to the th slope treatment unit in the landslide-prone area rectification concentration, represents a constant.

[0169] S6. Conduct risk division for the area to be measured based on the corrected landslide disaster susceptibility evaluation results. According to the corrected landslide disaster susceptibility evaluation results, divide the area to be measured by the natural breaks method to obtain the corresponding risk division results, as shown in Table 1. The risk division results include high risk, relatively high risk, medium risk, relatively low risk, and low risk.

[0170] Table 1

[0171]

[0172] As Figure 2 shown, a dynamic landslide disaster risk assessment system based on the progress of treatment projects includes:

[0173] Slope unit construction module, used to construct the slope unit set and the landslide disaster global susceptibility evaluation data set of the area to be measured, and divide the positive sample set of landslide disaster hidden danger points;

[0174] Pseudo-negative sample identification and generation training module, used to generate a pseudo-negative sample identification training set by the SPY-FPS pseudo-negative sample screening method, and train a classifier to obtain a pseudo-negative sample identification generator;

[0175] Pseudo-negative sample identification and acquisition module, used to input the landslide disaster global susceptibility evaluation data set into the pseudo-negative sample identification generator to obtain a pseudo-negative sample set;

[0176] Landslide disaster susceptibility evaluation module, including a stacked generalization integration model training module and a landslide disaster susceptibility evaluation and prediction module;

[0177] Stacked generalization integration model training module, used to input the landslide disaster susceptibility evaluation training data set into the stacked generalization integration model, and train it using the Bayesian hyperparameter optimization method guided by information entropy;

[0178] The landslide disaster susceptibility evaluation and prediction module is used to input the combined identified pseudo-negative sample set and the positive sample set of landslide disaster hidden danger points into the stacked generalization integration model to obtain the landslide disaster susceptibility evaluation result;

[0179] The landslide disaster susceptibility evaluation result correction module is used to obtain the landslide disaster treatment progress data in real time, construct a susceptibility correction coefficient and correct the landslide disaster susceptibility evaluation result;

[0180] The landslide risk division module is used to divide the risk of the area to be measured based on the corrected landslide disaster susceptibility evaluation result.

[0181] In summary, the present invention realizes the accurate and dynamic evaluation of landslide disaster risks by integrating multi-source heterogeneous data, real-time dynamic evaluation, and machine learning models, improves the evaluation accuracy and stability, and also enhances the timeliness and pertinence of decision-making; introduces the "feature perturbation stability" mechanism into the SPY (Spy Sample-Based Semi-Supervised Learning Algorithm) method to construct a more robust and discriminative pseudo-negative sample screening strategy, enabling effective identification of reliable negative samples in the data context of extremely unbalanced positive and negative samples; introduces the information entropy reduction amount to improve the acquisition function, enhances the efficiency of hyperparameter optimization, and effectively trains the stacked generalization integration model; considers the influence of the treatment progress in the landslide disaster area, sets a susceptibility correction coefficient, and corrects the landslide disaster susceptibility evaluation result to ensure that the landslide disaster susceptibility evaluation result can timely reflect the actual situation, providing strong technical support for the prevention and management of landslide disasters.

[0182] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, and all should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A dynamic risk assessment method for landslide disasters based on the progress of treatment projects, characterized in that, Including: Construct a slope unit set for the area to be measured and a dataset for the overall susceptibility assessment of landslide disasters, and divide the positive sample set of potential landslide disaster points; Use the SPY-FPS pseudo-negative sample screening method to generate a training set for identifying pseudo-negative samples, and train a classifier to obtain a pseudo-negative sample generator for identification; Input the dataset for the overall susceptibility assessment of landslide disasters into the pseudo-negative sample generator for identification to obtain a set of pseudo-negative samples for identification; Merge the set of pseudo-negative samples for identification and the positive sample set of potential landslide disaster points, and input them into a stacked generalization integration model to obtain the evaluation result of landslide disaster susceptibility; Obtain the landslide disaster treatment progress data in real time, construct a susceptibility correction coefficient, and correct the evaluation result of landslide disaster susceptibility; Based on the corrected evaluation result of landslide disaster susceptibility, conduct risk division for the area to be measured.

2. The dynamic risk assessment method for landslide disasters based on the progress of improvement projects according to claim 1, characterized in that, Construct a slope unit set for the area to be measured and a dataset for the overall susceptibility assessment of landslide disasters, and divide the positive sample set of potential landslide disaster points, including: Use the slope unit extraction method with forward and reverse hydrological division and terrain curvature constraint to process the area to be measured to obtain a slope unit set; Collect the initial susceptibility assessment data corresponding to the slope unit set, and connect the initial susceptibility assessment data and the slope unit set through a spatial join tool to obtain the initial dataset for the overall susceptibility assessment of landslide disasters; the initial susceptibility assessment data includes topographic data, geological data, meteorological data, and human activity data for each slope unit; Use the Spearman rank correlation coefficient to process the initial dataset for the overall susceptibility assessment of landslide disasters to obtain the positive sample set of potential landslide disaster points and the dataset for the overall susceptibility assessment of landslide disasters.

3. A dynamic risk assessment method for landslide disasters based on the progress of rectification projects according to claim 2, characterized in that, The process of obtaining the slope unit set is as follows: Collect high-precision DEM data of the area to be measured and input it into a geographic information system for depression filling processing to construct a high-precision digital elevation model; Calculate the flow direction and flow accumulation of each grid in the high-precision digital elevation model; Based on the flow accumulation, screen each grid to obtain the screened grids; According to the flow direction of each screened grid, use the watershed tool to divide the catchment area to generate a forward division result; Perform inversion processing on the high-precision DEM data to obtain high-precision inverted DEM data; perform processing on the high-precision inverted DEM data to generate a reverse division result; Use the union tool to merge the boundaries and unify the space of the forward division result and the forward division result to generate a slope boundary framework; Calculate the topographic profile curvature and planar curvature of the slope boundary framework and perform regional elimination to obtain the slope boundary framework after elimination; Use the polygon clipping tool to divide the area in the slope boundary framework after elimination to form a slope unit set.

4. A dynamic risk assessment method for landslide disasters based on the progress of rectification projects according to claim 1, characterized in that The training process of the classifier includes: Obtain the training dataset for the overall susceptibility assessment of landslide disasters, that is, obtain a positive sample training set and an unlabeled sample training set; Randomly select several positive samples from the positive sample training set and use them as an embedded reference positive sample training set; use the unselected positive samples in the positive sample training set as a new positive sample training set; Merge the embedded reference positive sample training set and the unlabeled sample training set to obtain a training set for identifying pseudo-negative samples; Input the training set of identified pseudo-negative samples and the training set of new positive samples into the classifier for training respectively to obtain an initially trained classifier; Obtain the classification probabilities of the embedded reference positive sample training set through the initially trained classifier, and set the confidence threshold; Based on the confidence threshold and the unlabeled sample training set, use the feature perturbation stability mechanism to generate a reliable negative sample training set, and input it into the initially trained classifier for secondary training to obtain a trained classifier, that is, an identified pseudo-negative sample generator.

5. A dynamic risk assessment method for landslide disasters based on the progress of improvement projects according to claim 1, characterized in that The stacked generalization ensemble model includes a base model and a meta-model connected in series; the base model includes decision trees, random forests, support vector machines, and gradient boosting trees that are processed in parallel; the meta-model is a meta-learner; Train the stacked generalization ensemble model using the Bayesian hyperparameter optimization method guided by information entropy.

6. The dynamic risk assessment method for landslide disasters based on the progress of improvement projects according to claim 5, characterized in that, The Bayesian hyperparameter optimization method guided by information entropy includes: Obtain the training set of identified pseudo-negative samples and the positive sample training set of landslide hazard hidden points and input them into the stacked generalization ensemble model; Define the hyperparameter combination of the stacked generalization ensemble model, construct the corresponding hyperparameter space; set the objective function; Construct a learnable surrogate model and set the information entropy acquisition function; determine the optimal hyperparameter combination through the learnable surrogate model and the information entropy acquisition function; Apply the optimal hyperparameter combination to the stacked generalization ensemble model to complete the training of the stacked generalization ensemble model.

7. A dynamic risk assessment method for landslide disasters based on the progress of rectification projects according to claim 6, characterized in that The process of determining the optimal hyperparameter combination is: Randomly select at least one hyperparameter combination and calculate the corresponding objective function value; Based on the selected hyperparameter combination and the objective function value, use Gaussian process regression to construct a learnable surrogate model; Set the information entropy acquisition function through the learnable surrogate model; calculate the information entropy acquisition function values of all hyperparameter combinations; Select the hyperparameter combination whose information entropy acquisition function value meets the screening conditions as the evaluation point for the next iteration; Apply the evaluation point to the stacked generalization ensemble model, calculate the corresponding objective function value, and update the learnable surrogate model; Repeat the selection of hyperparameter combinations and calculate the corresponding objective function values, and update the learnable surrogate model until the current iteration number reaches the maximum iteration number or the objective function value is less than the objective threshold to obtain the optimal hyperparameter combination.

8. A dynamic risk assessment method for landslide disasters based on the progress of rectification projects according to claim 7, characterized in that, The formula corresponding to the information entropy acquisition function is: ； ； Among them, and represent the second original acquisition function and the first original acquisition function respectively, represents the hyperparameter combination, represents the information entropy reduction, represents the optimal loss value, and represent the variance function and the mean function respectively, represents the intermediate parameter, and represent the standard normal cumulative distribution function and the standard normal probability density function respectively, represents the logarithmic function with a constant base, and represent the variance in the previous iteration and the variance in the current iteration respectively.

9. A dynamic risk assessment method for landslide disasters based on the progress of improvement projects according to claim 1, characterized in that The real-time acquisition of landslide disaster treatment progress data and the construction of a vulnerability correction coefficient to correct the landslide disaster vulnerability evaluation results include: Obtain the village landslide treatment points and the progress of the treatment situation in the area to be measured; Input the village landslide treatment points into the geographic information system for attribute connection to construct a landslide disaster treatment data set; Perform a spatial connection on the landslide disaster treatment data set and the landslide disaster vulnerability evaluation results to obtain a landslide-prone area treatment set; According to the treatment progress of each slope treatment unit in the landslide-prone area treatment set, set the engineering treatment progress percentage as the corresponding vulnerability correction coefficient; Based on the vulnerability correction coefficient, update the landslide disaster vulnerability evaluation results corresponding to each slope treatment unit in the landslide-prone area treatment set.

10. A landslide disaster risk dynamic assessment system based on the progress of rectification projects, which is used to implement the landslide disaster risk dynamic assessment method according to any one of claims 1 to 9, and is characterized in that, Include: The slope unit construction module is used to construct a set of slope units in the area to be measured and a dataset for the overall susceptibility evaluation of landslide disasters, and divide the positive sample set of potential landslide disaster points; The pseudo-negative sample identification and generation training module is used to generate a pseudo-negative sample identification training set by using the SPY-FPS pseudo-negative sample screening method, and train a classifier to obtain a pseudo-negative sample identification generator; The pseudo-negative sample identification and acquisition module is used to input the dataset for the overall susceptibility evaluation of landslide disasters into the pseudo-negative sample identification generator to obtain a pseudo-negative sample identification set; The landslide disaster susceptibility evaluation module includes a stacked generalization ensemble model training module and a landslide disaster susceptibility evaluation and prediction module; The stacked generalization ensemble model training module is used to input the landslide disaster susceptibility evaluation training dataset into the stacked generalization ensemble model and train it using the Bayesian hyperparameter optimization method guided by information entropy; The landslide disaster susceptibility evaluation and prediction module is used to input the combined pseudo-negative sample identification set and the positive sample set of potential landslide disaster points into the stacked generalization ensemble model to obtain the landslide disaster susceptibility evaluation result; The landslide disaster susceptibility evaluation result correction module is used to obtain the landslide disaster treatment progress data in real time, construct a susceptibility correction coefficient and correct the landslide disaster susceptibility evaluation result; The landslide risk division module is used to divide the risk of the area to be measured based on the corrected landslide disaster susceptibility evaluation result.

Images