Slope safety assessment method and system based on improved support vector regression

Abstract

The invention relates to the field of slope safety assessment, in particular to a slope safety assessment method and system based on improved support vector regression. A slope safety assessment system based on improved support vector regression comprises a data acquisition module, a slope safety assessment module, a feature importance assessment module and a decision module. According to the method, the improved support vector regression model is adopted, and the grid search and cross validation methods are combined to optimize the hyper-parameters, so that the safety coefficient of the slope can be predicted more accurately, and the accuracy and reliability of slope safety assessment are improved; through systematic hyper-parameter adjustment and verification, the generalization performance of the model on different data sets is ensured; besides, by calculating importance scores of the features, the method can identify and emphasize factors which have the greatest influence on the slope stability, so that deeper insight is provided for engineers to make more intelligent decisions.

Description

Technical Field

[0001] The invention relates to the field of slope safety assessment, and in particular to a slope safety assessment method and system based on improved support vector regression. Background Art

[0002] Slope safety assessment is a key task in the fields of civil engineering and geological engineering, and is of great significance for preventing natural disasters and ensuring the safety of human life and property. Traditional slope stability assessment methods mainly rely on analytical solutions and numerical simulation methods. These methods usually require complex mechanical models and a large amount of computing resources, and have extremely high requirements for parameter accuracy, which are often difficult to meet in practical applications. Analytical solutions cannot fully capture the complexity of slope behavior due to their inherent simplified assumptions, especially when the sliding surface is uncertain. Numerical simulation methods provide more accurate simulations, but require a large amount of in-situ measurement data and computing resources, and may be difficult to implement in some cases.

[0003] With the development of computer science, the application of machine learning methods in slope stability assessment has gradually attracted attention; machine learning models, especially support vector machine regression (SVR), have been widely studied due to their advantages in dealing with small samples and nonlinear problems; however, most existing studies focus on the development and verification of a single model, ignoring the importance of model generalization ability and hyperparameter optimization; in order to overcome the limitations of the prior art, the present invention proposes a slope safety assessment method and system based on improved support vector regression; the method optimizes hyperparameters through grid search and cross-validation, thereby improving the generalization ability and prediction accuracy of the model; at the same time, a feature importance scoring mechanism is introduced, and the overall change index is calculated through the feature importance scoring mechanism, which enhances the interpretability of the model and the ability to identify factors affecting slope stability, thereby realizing real-time monitoring of slope stability changes, thereby timely warning of potential slope instability risks. Summary of the invention

[0004] The present invention adopts an improved support vector regression model and combines grid search and cross-validation methods to optimize hyperparameters, which can more accurately predict the safety factor of the slope and improve the accuracy and reliability of slope safety assessment. Through systematic hyperparameter adjustment and verification, the generalization performance of the model on different data sets is ensured. In addition, by calculating the importance score of the features, the present invention can identify and emphasize the factors that have the greatest impact on slope stability, thereby providing engineers with deeper insights to make more informed decisions.

[0005] A slope safety assessment method based on improved support vector regression includes:

[0006] Collect characteristic data of the slope at any monitoring time point, the characteristic data include bulk density, soil cohesion, friction angle, slope angle, slope height and pore pressure ratio; standardize the collected characteristic data to obtain standard characteristic data, the standard characteristic data include standard value of bulk density, standard value of soil cohesion, standard value of friction angle, standard value of slope angle, standard value of slope height and standard value of pore pressure ratio;

[0007] A slope safety assessment model is constructed, wherein the hyperparameter combination of the slope safety assessment model is obtained through a grid search method, and the hyperparameters include penalty parameters and kernel function parameters; and the importance score of each feature in the feature data is calculated and obtained respectively;

[0008] The acquired standard characteristic data is used as the input of the slope safety assessment model, and the predicted slope safety factor is output;

[0009] After acquiring the standard feature data at the current monitoring time point, the overall change index is calculated based on all the historical standard feature data applied when the predicted slope safety factor was obtained most recently, and the importance score of each feature in the feature data is applied; it is determined whether the obtained overall change index is greater than the preset change index threshold. If so, the currently acquired standard feature data is used as the input of the slope safety assessment model, and the latest predicted slope safety factor is output; if not, no operation is performed.

[0010] Preferably, the hyperparameter combination of the slope safety assessment model is obtained by a grid search method, and the specific operation is as follows:

[0011] Set the value ranges for the penalty parameter C and the kernel function parameter g of the slope safety assessment model respectively. A pair of penalty parameter C and kernel function parameter g is a set of hyperparameter combinations. Combine all possible hyperparameter values ​​together to obtain a grid of hyperparameter combinations.

[0012] For any hyperparameter combination, cross validation is performed. The specific operations of cross validation include: dividing the training data set into k subsets of equal size, selecting one subset as the optimized validation set, and using the remaining k-1 subsets as the optimized training set; using the optimized training set to train the slope safety assessment model, and using the optimized validation set to evaluate the slope safety assessment model, and obtaining the evaluation indicators of the slope safety assessment model, which include determination coefficient, mean absolute percentage error, mean absolute error, and root mean square error;

[0013] For any hyperparameter combination, k rounds of cross validation are repeated, each time using a different optimized validation set, and the average value of the evaluation index of the slope safety assessment model in each round of k cross validation is calculated to obtain the final evaluation result of the current hyperparameter combination;

[0014] Compare the final evaluation results of each hyperparameter combination, select the optimal hyperparameter combination, and determine the penalty parameter C and kernel function parameter g of the slope safety assessment model.

[0015] Preferably, the slope safety assessment model is established based on the SVR model, the acquired standard characteristic data is used as the input of the slope safety assessment model, and the predicted slope safety factor is output. The specific operation is as follows;

[0016] The obtained standard values ​​of bulk density, soil cohesion, friction angle, slope angle, slope height and pore pressure ratio are combined into a characteristic vector x new ;

[0017] Using the formula Calculate and obtain the eigenvector x new With support vector x i The similarity between them, that is, K(x new ,x i ), where the support vectors are obtained during the training process, x i represents the i-th support vector, i = 1, 2, ..., n; n represents the number of support vectors; g is the kernel function parameter obtained by grid search; ‖x new -x ii ‖ 2 represents the feature vector x new and the support vector x i The square of the Euclidean distance between them;

[0018] Finally, using the formula Calculate the predicted value f(x new ), which is the predicted slope safety factor output by the slope safety assessment model; where α i represents the Lagrange multiplier of the i-th support vector, support vector x i The Lagrange multiplier of is obtained during the training process of the slope safety assessment model; b is the bias term obtained during the training process of the slope safety assessment model.

[0019] Preferably, the specific operations for training the slope safety assessment model are as follows:

[0020] Obtain a training data set, which includes several groups of training data, one group of training data includes standard values ​​of bulk density, standard values ​​of soil cohesion, standard values ​​of friction angle, standard values ​​of slope angle, standard values ​​of slope height, standard values ​​of pore pressure ratio and corresponding actual slope safety factors; divide the training data set into a training set and a validation set;

[0021] The obtained optimal hyperparameter combination is applied as the hyperparameter of the slope safety assessment model; the standard value of bulk density, standard value of soil cohesion, standard value of friction angle, standard value of slope angle, standard value of slope height and standard value of pore pressure ratio are used as the input of the slope safety assessment model, and the actual slope safety factor is used as the target output of the slope safety assessment model. The slope safety assessment model with parameter initialization is trained using the training set, and the slope safety assessment model is verified through the verification set to obtain the verification result; it is judged whether the verification result meets the preset training conditions. If so, the trained slope safety assessment model is output; if not, the slope safety assessment model continues to be trained through the training set.

[0022] Preferably, the importance score of each feature in the feature data is calculated separately, and the specific operations are as follows:

[0023] Randomly extract m evaluation samples from the training data set, each of which contains a set of standard values ​​for bulk density, soil cohesion, friction angle, slope angle, slope height, pore pressure ratio, and the corresponding actual slope safety factor; input the m evaluation samples into the slope safety assessment model in turn to obtain m predicted slope safety factors y j , j = 1, 2, ..., m; based on the actual slope safety factor in m evaluation samples Using the formula Calculate the original prediction error RMSE original ;

[0024] For any one of the characteristics of bulk density, soil cohesion, friction angle, slope angle, slope height and pore pressure ratio, the standard value corresponding to the characteristic in the m evaluation samples is randomly shuffled to obtain m shuffled samples of the current characteristic; the m shuffled samples of the current characteristic are sequentially input into the slope safety assessment model to obtain m predicted slope safety factors y corresponding to the current characteristic. l,j , l = 1, 2, ..., 6, which correspond to six features respectively; then use the formula Calculate the perturbation prediction error RMSE of the current feature l,perturbed ;

[0025] For any feature, based on the original prediction error RMSE original , using the formula ΔRMSE l =|RMSE l,perturbed -RMSE original |Calculate the error change of the current feature; finally use the formula Calculate the importance score of the current feature.

[0026] Preferably, the overall change index is calculated and obtained, and the specific operations are as follows:

[0027] Get the latest standard value of bulk density x γ,new , the latest soil cohesion standard value x c,new , the latest standard value of friction angle Latest standard value of slope angle x a,new , the latest standard value of slope height x H,new and the latest standard value of pore pressure ratio x r,new ; Based on the standard value of bulk density x used in the most recent prediction of slope safety factor γ , soil cohesion standard value x c , friction angle standard value Standard value of slope angle x a , standard value of slope height x H and the standard value of pore pressure ratio x r ; Calculate the difference respectively and obtain the bulk density change value d γ , soil cohesion change value d c , Friction angle change value Slope angle change d a , slope height change value d H and the change in pore pressure ratio d r ;

[0028] Using the formula Calculate and obtain the overall change index; ω 1 ,ω 2 ,ω 3 ,ω 4 ,ω 5 and ω 6 The importance scores are given for bulk density, soil cohesion, friction angle, slope angle, slope height and pore pressure ratio.

[0029] Preferably, the collected feature data is standardized to obtain standard feature data, and the specific operations are as follows:

[0030] Based on the obtained bulk density γ, soil cohesion c, friction angle The slope angle a, slope height H and pore pressure ratio r are calculated using the formula Calculate the standard value of bulk density x γ , where γ max and γ min are the preset maximum and minimum bulk density values ​​respectively; Calculate the standard value x of soil cohesion c , where c max and c min are the preset maximum soil cohesion and minimum soil cohesion respectively; Calculate the standard value of friction angle in and are the preset maximum friction angle and minimum friction angle respectively; Calculate the standard value of the slope angle x a , where a max and a min are the preset maximum slope angle and minimum slope angle respectively; Calculate the standard value of slope height x H , where H max and H min are the preset maximum slope height and minimum slope height respectively; Calculate the standard value x of pore pressure ratio r , where r max and r min are the preset maximum pore pressure ratio and minimum pore pressure ratio respectively.

[0031] Preferably, a slope safety assessment system based on improved support vector regression comprises:

[0032] The data acquisition module includes a data acquisition unit and a data processing unit; the data acquisition unit is used to collect characteristic data of the slope at any monitoring time point; the data processing unit is used to perform standard processing on the collected characteristic data to obtain standard characteristic data;

[0033] A slope safety assessment module is used to use the acquired standard characteristic data as the input of the slope safety assessment model and output the predicted slope safety factor;

[0034] The feature importance evaluation module is used to calculate the importance score of each feature in the feature data;

[0035] The decision-making module is used to obtain the standard feature data at any monitoring time point, and then calculate the overall change index based on all the historical standard feature data applied when the predicted slope safety factor was obtained most recently, and apply the importance score of each feature in the feature data to obtain the overall change index; determine whether the obtained overall change index is greater than the preset change index threshold. If so, use the currently obtained standard feature data as the input of the slope safety assessment model and output the latest predicted slope safety factor; if not, no operation is performed.

[0036] The present invention has the following advantages:

[0037] 1. The present invention adopts an improved support vector regression model and combines grid search and cross-validation methods to optimize hyperparameters, which can more accurately predict the safety factor of the slope and improve the accuracy and reliability of slope safety assessment. Through systematic hyperparameter adjustment and verification, the generalization performance of the model on different data sets is ensured. In addition, by calculating the importance score of the features, the present invention can identify and emphasize the factors that have the greatest impact on slope stability, thereby providing engineers with deeper insights to make more informed decisions.

[0038] 2. The present invention can monitor the changes in key parameters of slope stability in real time by calculating the overall change index and comparing it with a preset threshold. When a significant change is detected, the system will automatically trigger a reassessment and timely update the prediction of the slope safety factor, thereby providing an early warning for possible slope instability. This active monitoring and early warning mechanism significantly improves the ability to respond to slope safety risks, helps prevent disasters, and protects people’s lives and property. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic diagram of the structure of a slope safety assessment system based on improved support vector regression adopted in an embodiment of the present invention;

[0040] Figure 2 It is a schematic diagram of the framework of the slope safety assessment model adopted in the embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0042] Embodiment 1, a slope safety assessment method based on improved support vector regression, comprising:

[0043] The characteristic data of the slope are collected at any monitoring time point. The characteristic data include bulk density, soil cohesion, friction angle, slope angle, slope height and pore pressure ratio; bulk density refers to the weight of the soil per unit volume, which affects the stability of the slope; soil cohesion is the inherent cohesion between soil particles, which has an important influence on the shear strength of the slope; friction angle is the ability of the soil to resist shear failure, which is an important parameter for evaluating the stability of the slope; slope angle is the inclination angle of the slope, which is directly related to the geometric stability of the slope; slope height is the vertical height of the slope, which affects the development and stability of the potential slip surface; pore pressure ratio is the ratio of the pore water pressure in the soil to the soil stress, which affects The effective stress and stability of soil; standardize the collected characteristic data to obtain standard characteristic data. Standardization ensures the comparability between different characteristics, eliminates the impact of different dimensions and magnitudes, and helps to improve the efficiency of model training and the accuracy of prediction. The standard characteristic data include standard values ​​of bulk density, soil cohesion, friction angle, slope angle, slope height and pore pressure ratio. This process not only provides the necessary input data for the model, but also enhances the versatility of the data and the adaptability of the model through standardization, laying a solid foundation for accurately predicting the safety factor of the slope.

[0044] Construct slope safety assessment model, such as Figure 2 As shown in the figure, the hyperparameter combination of the slope safety assessment model is obtained through the grid search method, and the hyperparameters include penalty parameters and kernel function parameters; the penalty parameter controls the degree of penalty for misclassification of the model, and a larger penalty parameter value will make the model pay more attention to correctly classifying all training data, which may lead to overfitting; and the kernel function parameter affects the complexity of the kernel function, which is used to map the input data to a high-dimensional space; at the same time, the importance score of each feature in the feature data is calculated separately; the feature importance score helps to understand which features are most critical to the prediction of the slope safety factor, which not only helps to optimize the model, but also can provide guidance for engineering practice;

[0045] The acquired standard characteristic data is used as the input of the slope safety assessment model, and the predicted slope safety factor is output;

[0046] After acquiring the standard feature data at the current monitoring time point, the overall change index is calculated based on all the historical standard feature data applied when the predicted slope safety factor was obtained most recently, and the importance score of each feature in the feature data is applied; the overall change index provides a quantitative indicator for evaluating the change in slope stability and providing a basis for subsequent decision-making, and different weights are assigned taking into account the impact of different features on the prediction results, so that when the features with higher importance scores change, the overall change index can change significantly; it is judged whether the acquired overall change index is greater than the preset change index threshold. If so, it indicates that the slope characteristics have changed significantly and may affect the stability of the slope. The currently acquired standard feature data is used as the input of the slope safety assessment model, and the latest predicted slope safety factor is output; if not, it indicates that the slope characteristics have not changed significantly, and the system does not need to be re-evaluated, that is, no operation is performed; this decision-making mechanism ensures that the model is re-evaluated only when necessary, which improves the efficiency and response speed of the assessment, while avoiding unnecessary calculations and possible false alarms.

[0047] The hyperparameter combination of the slope safety assessment model is obtained through the grid search method. The specific operations are as follows:

[0048] Set the value ranges for the penalty parameter C and the kernel function parameter g of the slope safety assessment model respectively. A pair of penalty parameter C and kernel function parameter g is a set of hyperparameter combinations. Combine all possible hyperparameter values ​​together to obtain a grid of hyperparameter combinations.

[0049] For any hyperparameter combination, cross validation is performed. The specific operations of cross validation include: dividing the training data set into k subsets of equal size, selecting one subset as the optimized validation set, and using the remaining k-1 subsets as the optimized training set; using the optimized training set to train the slope safety assessment model, and using the optimized validation set to evaluate the slope safety assessment model, and obtaining the evaluation indicators of the slope safety assessment model, which include determination coefficient, mean absolute percentage error, mean absolute error, and root mean square error;

[0050] For any hyperparameter combination, k rounds of cross validation are repeated, each time using a different optimized validation set, and the average value of the evaluation index of the slope safety assessment model in each round of k cross validation is calculated to obtain the final evaluation result of the current hyperparameter combination;

[0051] Compare the final evaluation results of each hyperparameter combination, select the optimal hyperparameter combination, and determine the penalty parameter C and kernel function parameter g of the slope safety assessment model.

[0052] The slope safety assessment model is established based on the SVR model. The acquired standard characteristic data is used as the input of the slope safety assessment model, and the predicted slope safety factor is output. The specific operations are as follows;

[0053] The obtained standard values ​​of bulk density, soil cohesion, friction angle, slope angle, slope height and pore pressure ratio are combined into a characteristic vector x new ;

[0054] Using the formula Calculate and obtain the eigenvector x new With support vector x i The similarity between them, that is, K(x new ,x i ), where the support vectors are obtained during the training process, x i represents the i-th support vector, i = 1, 2, ..., n; n represents the number of support vectors; g is the kernel function parameter obtained by grid search; ‖x new -x ii ‖ 2 represents the feature vector x new and the support vector x i The square of the Euclidean distance between them;

[0055] Finally, using the formula Calculate the predicted value f(x new ), which is the predicted slope safety factor output by the slope safety assessment model; where α i represents the Lagrange multiplier of the i-th support vector, support vector x i The Lagrange multiplier of is obtained during the training process of the slope safety assessment model; b is the bias term obtained during the training process of the slope safety assessment model.

[0056] The specific operations for the training of the slope safety assessment model are as follows:

[0057] Obtain a training data set, which includes several groups of training data, one group of training data includes standard values ​​of bulk density, standard values ​​of soil cohesion, standard values ​​of friction angle, standard values ​​of slope angle, standard values ​​of slope height, standard values ​​of pore pressure ratio and corresponding actual slope safety factors; divide the training data set into a training set and a validation set;

[0058] The obtained optimal hyperparameter combination is applied as the hyperparameter of the slope safety assessment model; the standard value of bulk density, standard value of soil cohesion, standard value of friction angle, standard value of slope angle, standard value of slope height and standard value of pore pressure ratio are used as the input of the slope safety assessment model, and the actual slope safety factor is used as the target output of the slope safety assessment model. The slope safety assessment model with parameter initialization is trained using the training set, and the slope safety assessment model is verified through the verification set to obtain the verification result; it is judged whether the verification result meets the preset training conditions. If so, the trained slope safety assessment model is output; if not, the slope safety assessment model continues to be trained through the training set.

[0059] Calculate the importance score of each feature in the feature data separately. The specific operations are as follows:

[0060] Randomly extract m evaluation samples from the training data set, each of which contains a set of standard values ​​for bulk density, soil cohesion, friction angle, slope angle, slope height, pore pressure ratio, and the corresponding actual slope safety factor; input the m evaluation samples into the slope safety assessment model in turn to obtain m predicted slope safety factors y j , j = 1, 2, ..., m; based on the actual slope safety factor in m evaluation samples Using the formula Calculate the original prediction error RMSE original ;

[0061] For any one of the characteristics of bulk density, soil cohesion, friction angle, slope angle, slope height and pore pressure ratio, the standard value corresponding to the characteristic in the m evaluation samples is randomly shuffled to obtain m shuffled samples of the current characteristic; the m shuffled samples of the current characteristic are sequentially input into the slope safety assessment model to obtain m predicted slope safety factors y corresponding to the current characteristic. l,j , l = 1, 2, ..., 6, which correspond to six features respectively; then use the formula Calculate the perturbation prediction error RMSE of the current feature l,perturbed ;

[0062] For any feature, based on the original prediction error RMSE original , using the formula ΔRMSE l =|RMSE l,perturbed -RMSE original |Calculate the error change of the current feature; finally use the formula Calculate the importance score of the current feature.

[0063] Calculate and obtain the overall change index. The specific operations are as follows:

[0064] Get the latest standard value of bulk density x γ,new , the latest soil cohesion standard value x c,new , the latest standard value of friction angle The latest standard value of slope angle x a,new , the latest standard value of slope height x H,new and the latest standard value of pore pressure ratio x r,new ; Based on the standard value of bulk density x used in the most recent prediction of slope safety factor γ , soil cohesion standard value x c , friction angle standard value Standard value of slope angle x a , standard value of slope height x H and the standard value of pore pressure ratio x r ; Calculate the difference respectively and obtain the bulk density change value d γ , soil cohesion change value d c , Friction angle change value Slope angle change value d a , slope height change value d H and the pore pressure ratio change d r ;

[0065] Using the formula Calculate and obtain the overall change index; ω 1 ,ω 2 ,ω 3 ,ω 4 ,ω 5 and ω 6 The importance scores are given for bulk density, soil cohesion, friction angle, slope angle, slope height and pore pressure ratio.

[0066] Standardize the collected feature data to obtain standard feature data. The specific operations are as follows:

[0067] Based on the obtained bulk density γ, soil cohesion c, friction angle The slope angle a, slope height H and pore pressure ratio r are calculated using the formula Calculate the standard value of bulk density x γ , where γ max and γ min are the preset maximum and minimum bulk density values ​​respectively; Calculate the standard value x of soil cohesion c , where c max and c min are the preset maximum soil cohesion and minimum soil cohesion respectively; Calculate the standard value of friction angle in and are the preset maximum friction angle and minimum friction angle respectively; Calculate the standard value of the slope angle x a , where a max and a min are the preset maximum slope angle and minimum slope angle respectively; Calculate the standard value of slope height x H , where H max and H min are the preset maximum slope height and minimum slope height respectively; Calculate the standard value x of pore pressure ratio r , where r max and r min are the preset maximum pore pressure ratio and minimum pore pressure ratio respectively.

[0068] Example 2, a slope safety assessment system based on improved support vector regression, such as Figure 1 As shown, including:

[0069] The data acquisition module includes a data acquisition unit and a data processing unit; the data acquisition unit is used to collect characteristic data of the slope at any monitoring time point; the data processing unit is used to perform standard processing on the collected characteristic data to obtain standard characteristic data;

[0070] A slope safety assessment module is used to use the acquired standard characteristic data as the input of the slope safety assessment model and output the predicted slope safety factor;

[0071] The feature importance evaluation module is used to calculate the importance score of each feature in the feature data;

[0072] The decision-making module is used to obtain the standard feature data at any monitoring time point, and then calculate the overall change index based on all the historical standard feature data applied when the predicted slope safety factor was obtained most recently, and apply the importance score of each feature in the feature data to obtain the overall change index; determine whether the obtained overall change index is greater than the preset change index threshold. If so, use the currently obtained standard feature data as the input of the slope safety assessment model and output the latest predicted slope safety factor; if not, no operation is performed.

[0073] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A slope safety assessment method based on improved support vector regression, characterized in that: include: Collect characteristic data of the slope at any monitoring time point, the characteristic data include bulk density, soil cohesion, friction angle, slope angle, slope height and pore pressure ratio; standardize the collected characteristic data to obtain standard characteristic data, the standard characteristic data include standard value of bulk density, standard value of soil cohesion, standard value of friction angle, standard value of slope angle, standard value of slope height and standard value of pore pressure ratio; A slope safety assessment model is constructed, wherein the hyperparameter combination of the slope safety assessment model is obtained through a grid search method, and the hyperparameters include penalty parameters and kernel function parameters; and the importance score of each feature in the feature data is calculated and obtained respectively; The acquired standard characteristic data is used as the input of the slope safety assessment model, and the predicted slope safety factor is output; After acquiring the standard feature data at the current monitoring time point, the overall change index is calculated based on all the historical standard feature data applied when the predicted slope safety factor was obtained most recently, and the importance score of each feature in the feature data is applied; it is determined whether the obtained overall change index is greater than the preset change index threshold. If so, the currently acquired standard feature data is used as the input of the slope safety assessment model, and the latest predicted slope safety factor is output; if not, no operation is performed.

2. A slope safety assessment method based on improved support vector regression according to claim 1, characterized in that: The hyperparameter combination of the slope safety assessment model is obtained through the grid search method. The specific operations are as follows: Set the value ranges for the penalty parameter C and the kernel function parameter g of the slope safety assessment model respectively. A pair of penalty parameter C and kernel function parameter g is a set of hyperparameter combinations. Combine all possible hyperparameter values ​​together to obtain a grid of hyperparameter combinations. For any hyperparameter combination, cross validation is performed. The specific operations of cross validation include: dividing the training data set into k subsets of equal size, selecting one subset as the optimized validation set, and using the remaining k-1 subsets as the optimized training set; using the optimized training set to train the slope safety assessment model, and using the optimized validation set to evaluate the slope safety assessment model, and obtaining the evaluation indicators of the slope safety assessment model, which include determination coefficient, mean absolute percentage error, mean absolute error, and root mean square error; For any hyperparameter combination, k rounds of cross validation are repeated, each time using a different optimized validation set, and the average value of the evaluation index of the slope safety assessment model in each round of k cross validation is calculated to obtain the final evaluation result of the current hyperparameter combination; Compare the final evaluation results of each hyperparameter combination, select the optimal hyperparameter combination, and determine the penalty parameter C and kernel function parameter g of the slope safety assessment model.

3. A slope safety assessment method based on improved support vector regression according to claim 2, characterized in that: The slope safety assessment model is established based on the SVR model. The acquired standard characteristic data is used as the input of the slope safety assessment model, and the predicted slope safety factor is output. The specific operations are as follows; The obtained standard values ​​of bulk density, soil cohesion, friction angle, slope angle, slope height and pore pressure ratio are combined into a characteristic vector x new ; Using the formula Calculate and obtain the eigenvector x new With support vector x i The similarity between them, that is, K(x new ,x i ), where the support vectors are obtained during the training process, x i represents the i-th support vector, i = 1, 2, ..., n; n represents the number of support vectors; g is the kernel function parameter obtained by grid search; ‖x new -x ii ‖ 2 represents the feature vector x new and the support vector x i The square of the Euclidean distance between them; Finally, using the formula Calculate the predicted value f(x new ), which is the predicted slope safety factor output by the slope safety assessment model; where α i represents the Lagrange multiplier of the i-th support vector, support vector x i The Lagrange multiplier of is obtained during the training process of the slope safety assessment model; b is the bias term obtained during the training process of the slope safety assessment model.

4. A slope safety assessment method based on improved support vector regression according to claim 3, characterized in that: The specific operations for the training of the slope safety assessment model are as follows: Obtain a training data set, which includes several groups of training data, one group of training data includes standard values ​​of bulk density, standard values ​​of soil cohesion, standard values ​​of friction angle, standard values ​​of slope angle, standard values ​​of slope height, standard values ​​of pore pressure ratio and corresponding actual slope safety factors; divide the training data set into a training set and a validation set; The obtained optimal hyperparameter combination is applied as the hyperparameter of the slope safety assessment model; the standard value of bulk density, standard value of soil cohesion, standard value of friction angle, standard value of slope angle, standard value of slope height and standard value of pore pressure ratio are used as the input of the slope safety assessment model, and the actual slope safety factor is used as the target output of the slope safety assessment model. The slope safety assessment model with parameter initialization is trained using the training set, and the slope safety assessment model is verified through the verification set to obtain the verification result; it is judged whether the verification result meets the preset training conditions. If so, the trained slope safety assessment model is output; if not, the slope safety assessment model continues to be trained through the training set.

5. A slope safety assessment method based on improved support vector regression according to claim 4, characterized in that: Calculate the importance score of each feature in the feature data separately. The specific operations are as follows: Randomly extract m evaluation samples from the training data set, each of which contains a set of standard values ​​for bulk density, soil cohesion, friction angle, slope angle, slope height, pore pressure ratio, and the corresponding actual slope safety factor; input the m evaluation samples into the slope safety assessment model in turn to obtain m predicted slope safety factors y j , j = 1, 2, ..., m; Actual slope safety factor based on m evaluation samples Using the formula Calculate the original prediction error RMSE original ; For any one of the characteristics of bulk density, soil cohesion, friction angle, slope angle, slope height and pore pressure ratio, the standard value corresponding to the characteristic in the m evaluation samples is randomly shuffled to obtain m shuffled samples of the current characteristic; the m shuffled samples of the current characteristic are sequentially input into the slope safety assessment model to obtain m predicted slope safety factors y corresponding to the current characteristic. l,j , l = 1, 2, ..., 6, indicating that they correspond to six features respectively; Then use the formula Calculate the perturbation prediction error RMSE of the current feature l,perturbed ; For any feature, based on the original prediction error RMSE original , using the formula ΔRMSE l =|RMSE l,perturbed -RMSE original |Calculate the error change of the current feature; Finally, using the formula Calculate the importance score of the current feature.

6. A slope safety assessment method based on improved support vector regression according to claim 5, characterized in that: Calculate and obtain the overall change index. The specific operations are as follows: Get the latest standard value of bulk density x γ,new , the latest soil cohesion standard value x c,new , the latest standard value of friction angle Latest standard value of slope angle x a,new , the latest standard value of slope height x H,new and the latest standard value of pore pressure ratio x r,new ; Based on the standard value of bulk density x used in the most recent prediction of slope safety factor γ , soil cohesion standard value x c , friction angle standard value Standard value of slope angle x a , standard value of slope height x H and the standard value of pore pressure ratio x r ; Calculate the difference respectively and obtain the bulk density change value d γ , soil cohesion change value d c , Friction angle change value Slope angle change d a , slope height change value d H and the change in pore pressure ratio d r ; Using the formula The overall variation index is calculated; ω1, ω2, ω3, ω4, ω5 and ω6 are the importance scores of bulk density, soil cohesion, friction angle, slope angle, slope height and pore pressure ratio, respectively.

7. A slope safety assessment method based on improved support vector regression according to claim 6, characterized in that: Standardize the collected feature data to obtain standard feature data. The specific operations are as follows: Based on the obtained bulk density γ, soil cohesion c, friction angle The slope angle a, slope height H and pore pressure ratio r are calculated using the formula Calculate the standard value of bulk density x γ , where γ max and γ min are the preset maximum and minimum bulk density values ​​respectively; Calculate the standard value x of soil cohesion c , where c max and c min are the preset maximum soil cohesion and minimum soil cohesion respectively; Calculate the standard value of friction angle in and are the preset maximum friction angle and minimum friction angle respectively; Calculate the standard value of the slope angle x a , where a max and a min are the preset maximum slope angle and minimum slope angle respectively; Calculate the standard value of slope height x H , where H max and H min are the preset maximum slope height and minimum slope height respectively; Calculate the standard value x of pore pressure ratio r , where r max and r min are the preset maximum pore pressure ratio and minimum pore pressure ratio respectively.

8. A slope safety assessment system based on improved support vector regression, characterized in that: The system is applied to a slope safety assessment method based on improved support vector regression as described in any one of claims 1 to 7, comprising: The data acquisition module includes a data acquisition unit and a data processing unit; the data acquisition unit is used to collect characteristic data of the slope at any monitoring time point; the data processing unit is used to perform standard processing on the collected characteristic data to obtain standard characteristic data; A slope safety assessment module is used to use the acquired standard characteristic data as the input of the slope safety assessment model and output the predicted slope safety factor; The feature importance evaluation module is used to calculate the importance score of each feature in the feature data; The decision-making module is used to obtain the standard feature data at any monitoring time point, and then calculate the overall change index based on all the historical standard feature data applied when the predicted slope safety factor was obtained most recently, and apply the importance score of each feature in the feature data to obtain the overall change index; determine whether the obtained overall change index is greater than the preset change index threshold. If so, use the currently obtained standard feature data as the input of the slope safety assessment model and output the latest predicted slope safety factor; if not, no operation is performed.

Images