Method and device for constructing a tunnel extrusion deformation prediction model

Abstract

The application provides a tunnel extrusion deformation prediction model construction method and device. The method comprises: obtaining case data of tunnel extrusion deformation to generate an initial data set; using multiple data filling methods to fill data in the initial data set to generate multiple complete data sets; training multiple first machine learning models according to the complete data sets, and determining model training data sets according to model performance indicators of the multiple first machine learning models; training multiple second machine learning models according to training sets in the model training data sets, and optimizing hyperparameters in the training process of the second machine learning models by using an improved black-winged kite algorithm; and determining a tunnel extrusion deformation prediction model according to model performance indicators of the multiple second machine learning models obtained by training. The model performance of the tunnel extrusion deformation prediction model obtained by training can be improved.

Description

Technical Field

[0001] This invention belongs to the field of tunnel engineering, and in particular relates to a method and apparatus for constructing a tunnel compression deformation prediction model. Background Technology

[0002] Due to the complexity of geological conditions, large-scale compression deformation disasters frequently occur during tunnel construction. In past engineering examples and projects, limitations imposed by construction sites and underground engineering conditions often prevent the direct acquisition of complete geotechnical mechanical characteristics. This results in a lack of comprehensive data records in these cases, making it difficult to establish accurate tunnel compression deformation prediction models. Consequently, it is challenging to provide experience and guidance for assessing the potential for large-scale compression deformation disasters in existing unexcavated tunnels and projects yet to be constructed. Summary of the Invention

[0003] This invention provides a method and apparatus for constructing a tunnel extrusion deformation prediction model, which can improve the model performance of the tunnel extrusion deformation prediction model.

[0004] On the one hand, a method for constructing a tunnel extrusion deformation prediction model is provided, including:

[0005] Obtain case data on tunnel compression deformation to generate an initial dataset;

[0006] Multiple data filling methods are used to fill the initial dataset to generate multiple complete datasets;

[0007] Train multiple first-order machine learning models based on the complete dataset, and determine the model training dataset based on the model performance metrics of the multiple first-order machine learning models.

[0008] Multiple second machine learning models are trained based on the training set in the model training dataset. During the model training process, the improved Black Kite algorithm is used to optimize the hyperparameters of the second machine learning models.

[0009] The tunnel compression deformation prediction model is determined based on the model performance metrics of multiple second machine learning models obtained through training.

[0010] Optionally, multiple data filling methods are used to fill the initial dataset to generate multiple complete datasets, including:

[0011] The initial dataset was filled using the mean imputation method, median imputation method, mode imputation method, multiple imputation method, and KNN imputation method respectively, resulting in multiple complete datasets.

[0012] Optionally, multiple first machine learning models are trained based on the complete dataset, and the model training dataset is determined based on the model performance metrics of the multiple first machine learning models, including:

[0013] For each complete dataset, multiple machine learning models are trained using different machine learning algorithms;

[0014] Calculate the sum of model performance metrics for the machine learning model, wherein the model performance metrics include any one of the following: weighted average of accuracy and recall, accuracy, Matthews correlation coefficient, and Kappa coefficient;

[0015] The machine learning model with the largest sum of its performance metrics is identified, and the complete dataset used for the model training dataset is the machine learning model with the largest sum of its performance metrics.

[0016] Optionally, multiple second machine learning models are trained based on the training set in the model training dataset. During the model training process, an improved Black-winged Kite algorithm is used to optimize the hyperparameters of the second machine learning models, including:

[0017] Using a weighted average of precision and recall as the iterative optimization objective, an improved black-winged kite algorithm is employed to optimize the hyperparameters during the training process of multiple machine learning models. The improved black-winged kite algorithm is obtained by using a logistic mapping to improve the initial population position of the black-winged kite algorithm. The initial population position of the improved black-winged kite algorithm is... ;

[0018] Based on the training set in the model training dataset, multiple second machine learning models are obtained by using the ten-fold cross-validation training method and multiple optimized machine learning model training algorithms.

[0019] Optionally, a tunnel extrusion deformation prediction model is determined based on the model performance metrics of multiple trained second machine learning models, including:

[0020] Based on the second machine learning model, the model performance metrics of the training set of the second machine learning model are obtained. The model performance metrics include any one of the following: weighted average of accuracy and recall, accuracy, Matthews correlation coefficient, and Kappa coefficient.

[0021] The second machine learning model is tested based on the test set to obtain the model performance metrics of the second machine learning model on the test set.

[0022] The model performance metrics of the second machine learning model are summed from the training and testing sets to obtain the sum of the model performance metrics of the second machine learning model. The second machine learning model with the largest sum of model performance metrics is the tunnel extrusion deformation prediction model.

[0023] On the other hand, an apparatus for constructing a tunnel extrusion deformation prediction model is provided, characterized in that it includes:

[0024] The acquisition module is used to acquire case data on tunnel extrusion deformation and generate an initial dataset;

[0025] The data filling module is used to fill the initial dataset with data using multiple data filling methods to generate multiple complete datasets.

[0026] The model training module is used to train multiple first machine learning models based on the complete dataset, and to determine the model training dataset based on the model performance metrics of the multiple first machine learning models; to train multiple second machine learning models based on the training set in the model training dataset, and to optimize the hyperparameters of the second machine learning models during the training process using the improved Black Kite algorithm; and to determine the tunnel extrusion deformation prediction model based on the model performance metrics of the multiple trained second machine learning models.

[0027] Optionally, the model training module is used for:

[0028] For each complete dataset, multiple machine learning models are trained using different machine learning algorithms;

[0029] Calculate the sum of model performance metrics for the machine learning model, wherein the model performance metrics include any one of the following: weighted average of accuracy and recall, accuracy, Matthews correlation coefficient, and Kappa coefficient;

[0030] The machine learning model with the largest sum of its performance metrics is identified, and the complete dataset used for the model training dataset is the machine learning model with the largest sum of its performance metrics.

[0031] Optionally, the model training module is used for:

[0032] Using a weighted average of precision and recall as the iterative optimization objective, an improved black-winged kite algorithm is employed to optimize the hyperparameters during the training process of multiple machine learning models. The improved black-winged kite algorithm is obtained by using a logistic mapping to improve the initial population position of the black-winged kite algorithm. The initial population position of the improved black-winged kite algorithm is... ;

[0033] Based on the training set in the model training dataset, multiple second machine learning models are obtained by using the ten-fold cross-validation training method and multiple optimized machine learning model training algorithms.

[0034] On the other hand, an electronic device is provided, including a construction apparatus for a tunnel extrusion deformation prediction model as described in any of the preceding claims.

[0035] On the other hand, a computer-readable storage medium is provided, wherein at least one piece of program code is stored in the computer-readable storage medium, the program code being executed by a processor to implement the method for constructing a tunnel extrusion deformation prediction model as described in any of the preceding claims.

[0036] The beneficial effects of the technical solutions provided in this disclosure are:

[0037] This disclosure provides a method for constructing a tunnel extrusion deformation prediction model. By acquiring case data and using multiple data imputation methods to fill the case data, multiple complete datasets are obtained, ensuring the completeness of the case data. After obtaining the complete datasets, different algorithms are used to train machine learning models based on each complete dataset. Then, the performance metrics of all machine learning models are evaluated to determine the optimal data, i.e., the data in the model training dataset, ensuring the performance of the subsequently trained tunnel extrusion deformation prediction model. After determining the model training dataset, multiple second machine learning models are trained based on the model training dataset, and the improved Black-winged Kite algorithm is used to optimize the hyperparameters during the model training process, ensuring the performance of the trained second machine learning models. After obtaining multiple second machine learning models, the performance metrics of all second machine learning models are statistically analyzed, and the second machine learning model with the best performance is selected as the tunnel extrusion deformation prediction model. The method for constructing a tunnel extrusion deformation prediction model provided by this invention ensures that the final trained tunnel extrusion deformation prediction model has good performance metrics. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0039] Figure 1 A flowchart illustrating a method for constructing a tunnel extrusion deformation prediction model, as provided in this embodiment of the disclosure;

[0040] Figure 2 A flowchart for initial dataset data filling provided in this embodiment of the disclosure;

[0041] Figure 3 A flowchart of an improved black-winged kite algorithm provided in this disclosure embodiment;

[0042] Figure 4 A flowchart illustrating the training process of a second machine learning model provided in this embodiment of the disclosure;

[0043] Figure 5 A structural block diagram of a device for constructing a tunnel extrusion deformation prediction model provided in this embodiment of the present disclosure;

[0044] Figure 6 This is a structural block diagram of an electronic device provided in an embodiment of the present disclosure.

[0045] The attached figures are labeled as follows:

[0046] 21: Acquisition module; 22: Data filling module; 23: Training module;

[0047] 31: Processor; 32: Memory. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0049] Variables typically used to predict large compression deformation disasters include: tunnel depth (H), tunnel diameter (D), support stiffness (K), rock mass quality index (Q), and stress-strength ratio (SSR). However, while tunnel deformation values ​​can be easily obtained from on-site surveys in existing engineering records, geotechnical characteristics such as Q, K, and SSR are not fully recorded. Therefore, it is not possible to analyze compression deformation disasters based on directly collected, incomplete datasets. It is necessary to first complete the data records and then establish a compression deformation prediction model based on a complete dataset. Hoek proposed a classification principle for large compression deformation based on tunnel compression deformation: let ε (%) represent the relative deformation of the tunnel; ε>10 indicates extremely severe compression deformation, 10>ε>5 indicates severe compression deformation, 5>ε>2.5 indicates moderate compression deformation, 2.5>ε>1 indicates slight compression deformation, and ε<1 indicates no compression deformation. Based on the variable characteristics and the classification principle for large compression deformation, a prediction model for large deformation levels was finally established.

[0050] Figure 1 A flowchart illustrating a method for constructing a tunnel compression deformation prediction model, as provided in this embodiment of the disclosure. See also... Figure 1 ,include:

[0051] S11. Obtain case data on tunnel compression deformation and generate an initial dataset.

[0052] In step S11, historical tunnel engineering compression deformation case data from various countries and regions are collected, including tunnel burial depth (H), tunnel diameter (D), support stiffness (K), rock mass quality index (Q), strength-stress ratio (SSR), and tunnel compression deformation level (Rank). Missing feature values ​​in the project are temporarily marked as Null, forming an incomplete dataset of tunnel engineering compression deformation disasters.

[0053] S12. Use multiple data filling methods to fill the initial dataset to generate multiple complete datasets.

[0054] In one example, step S12 includes:

[0055] The initial dataset was filled using the mean imputation method, median imputation method, mode imputation method, multiple imputation method, and KNN imputation method respectively, resulting in multiple complete datasets.

[0056] In this embodiment of the disclosure, the data in the complete dataset obtained by different data filling methods are different. By using multiple data filling methods to fill the data in the initial dataset, it can be ensured that the obtained complete dataset is relatively rich, which is beneficial for subsequent steps to determine the dataset most conducive to model training.

[0057] S13. Train multiple first machine learning models based on the complete dataset, and determine the model training dataset based on the model performance metrics of the multiple first machine learning models.

[0058] In one example, step S13 includes:

[0059] The first step is to train multiple machine learning models using different machine learning algorithms for each complete dataset.

[0060] In the first step, several machine learning algorithms are used, namely the Vector Machine (SVM) algorithm, the Random Forest (RF) algorithm, the Logistic Regression (LR) algorithm, the Adaptive Boosting (AdaBoost) algorithm, and the Extreme Boosting (XGBoost) algorithm.

[0061] Of course, it is not limited to the above algorithm; the above algorithm is just one example provided in this disclosure.

[0062] The second step is to calculate the sum of the model performance metrics of the machine learning model. The model performance metrics include any one of the following: the weighted average of accuracy and recall (F1), accuracy (ACC), Matthews correlation coefficient (MCC), and Kappa coefficient.

[0063] In this embodiment of the disclosure, the model performance metrics may not be limited to the model performance metrics described above.

[0064] The third step is to determine the machine learning model with the largest sum of model performance metrics. The complete dataset used for the machine learning model with the largest sum of model performance metrics is the model training dataset.

[0065] In this embodiment of the disclosure, each complete dataset is trained using multiple machine learning algorithms. After all complete datasets are trained, multiple first machine learning models are obtained. The model performance of the first machine learning models can be determined by summing the model performance metrics of the first machine learning models. The dataset used by the first machine learning model with the best model performance is most conducive to the training of the machine learning model.

[0066] S14. Train multiple second machine learning models based on the training set in the model training dataset. During the model training process, use the improved Blackwing Kite algorithm to optimize the hyperparameters of the second machine learning models.

[0067] In one example, step S14 includes:

[0068] The first step involves using F1 as the iterative optimization target and employing an improved black-winged kite algorithm to optimize the hyperparameters during the training process of multiple machine learning models. The improved black-winged kite algorithm is obtained by using a logistic mapping to improve the initial population position of the black-winged kite algorithm. The initial population position of the improved black-winged kite algorithm is... .

[0069] In the existing Black-winged Kite algorithm, the initial position is ,in, , They represent the first The upper and lower bounds of individual Black-winged Kites. It is a random number between [0,1].

[0070] In this embodiment of the disclosure, by introducing a Logistic mapping to modify the Black-winged Kite (BKA) algorithm to form the Improved Black-winged Kite (LBKA) algorithm, the richness of the algorithm population can be increased, and the possibility of escaping local optima can be increased.

[0071] The second step involves training multiple second machine learning models using the training set in the model training dataset and the ten-fold cross-validation training method and multiple optimized machine learning model training algorithms.

[0072] In the second step, the trained second machine learning models are LBKA-SVM, LBKA-RF, LBKA-LR, LBKA-AdaBoost, and LBKA-XGBoost, respectively, which are obtained by optimizing the hyperparameters of SVM, RF, LR, AdaBoost, and XGBoost using LBKA.

[0073] In this embodiment, the initial population position of the Black-winged Kite algorithm is improved by using Logistic mapping, forming an improved Black-winged Kite algorithm (denoted as LBKA). Modifying the initial population position of the Black-winged Kite algorithm through Logistic mapping, and then performing population attacks and migrations, increases the richness of population iterations and the convergence speed of the improved Black-winged Kite algorithm. Optimizing the hyperparameters during the training process of the machine learning model by improving the Black-winged Kite algorithm can enhance the performance of the trained machine learning model.

[0074] In this embodiment of the disclosure, the attack behavior of the improved Blackwing Kite Algorithm (LBKA) is as follows:

[0075] , ;

[0076] in, and Indicates the first Only the black-winged kite in the first Victor , The position of the next iteration step; A random number in [0,1] =0.9; This represents the total number of iterations. This represents the number of iterations completed so far.

[0077] The migration behavior of the improved Blackwing Kite Algorithm (LBKA) is as follows:

[0078] , ;

[0079] in, For the current number The iteration of the ... The leading scorer for the black-winged kite; For the first In the next iteration, any (i.e., any one) black-winged kite in the... Fitness value at a random location; For any (i.e., any) black-winged kite in the t-th iteration, in the t-th iteration... Fitness value of the dimensional position; It is a Cauchy mutation.

[0080] In this embodiment of the disclosure, the data in the model training dataset is divided into a training set and a test set, which are obtained by randomly selecting data from the model training dataset.

[0081] In one example, 80% of the data in the model training dataset is randomly selected as the training set, and the remaining 20% ​​is used as the test set.

[0082] Of course, the above is only one example of the selection of training and test set data, and this disclosure does not limit it.

[0083] S15. Determine the tunnel extrusion deformation prediction model based on the model performance indicators of multiple second machine learning models obtained from training.

[0084] In one example, step S15 includes:

[0085] The first step is to obtain the model performance metrics of the training set of the second machine learning model based on the second machine learning model. The model performance metrics include ACC, F1, Kappa, and MCC.

[0086] The second step is to test the second machine learning model on the test set to obtain the model performance metrics of the second machine learning model on the test set.

[0087] The third step is to sum the model performance metrics of the second machine learning model under the training set and the test set to obtain the sum of the model performance metrics of the second machine learning model. The second machine learning model with the largest sum of model performance metrics is the tunnel extrusion deformation prediction model.

[0088] In this embodiment, during model training, the improved Black-winged Kite algorithm is used to optimize the hyperparameters during training. Therefore, the performance of the final trained second machine learning model will differ somewhat from that of the first machine learning model. By evaluating the performance of the multiple final second machine learning models, the second machine learning model with the best performance is selected as the tunnel extrusion deformation prediction model.

[0089] This disclosure provides a method for constructing a tunnel extrusion deformation prediction model. By acquiring case data and using multiple data imputation methods to fill the case data, multiple complete datasets are obtained, ensuring the completeness of the case data. After obtaining the complete datasets, different algorithms are used to train machine learning models based on each complete dataset. Then, the performance metrics of all machine learning models are evaluated to determine the optimal data, i.e., the data in the model training dataset, ensuring the performance of the subsequently trained tunnel extrusion deformation prediction model. After determining the model training dataset, multiple second machine learning models are trained based on the model training dataset, and the improved Black-winged Kite algorithm is used to optimize the hyperparameters during the model training process, ensuring the performance of the trained second machine learning models. After obtaining multiple second machine learning models, the performance metrics of all second machine learning models are statistically analyzed, and the second machine learning model with the best performance is selected as the tunnel extrusion deformation prediction model. The method for constructing a tunnel extrusion deformation prediction model provided by this invention ensures that the final trained tunnel extrusion deformation prediction model has good performance metrics.

[0090] Figure 2 A flowchart illustrating an initial dataset data filling method provided for embodiments of this disclosure. See also... Figure 2 The initial dataset includes case data of various extrusion deformation levels. The initial dataset is filled using the mean imputation method, median imputation method, mode imputation method, multiple imputation method, and KNN imputation method, respectively, to obtain multiple complete datasets (stored in a complete database). The number of complete datasets corresponds to the number of data imputation methods used.

[0091] After the filling is completed, the model training algorithm in the classifier group is used to train multiple first machine learning models. The model performance of the first machine learning models is evaluated and summed to determine the final database (the final database stores the model training dataset).

[0092] Figure 3 A flowchart illustrating an improved black-winged kite algorithm provided in this disclosure. See also... Figure 3 The Improved Black-winged Kite Algorithm (LBKA) differs from the Black-winged Kite Algorithm (BKA) in that it modifies the initial population position of the Black-winged Kite Algorithm by altering the Logistic mapping. By modifying the initial population position of the Black-winged Kite Algorithm, the richness of population iteration and the speed of convergence are improved.

[0093] Figure 4 A flowchart illustrating the training process of a second machine learning model provided in an embodiment of this disclosure. See also... Figure 4 The second machine learning model is based on Figure 2The model is trained using a complete dataset (i.e., the model training dataset) as determined in the dataset. During training, the improved Black-winged Kite Algorithm (LBKA) is used to optimize the hyperparameters in the machine learning model training process. Figure 4 The hyperparameters optimized below SVM, RF, LR, AdaBoost, and XGBoost are shown in the table. After training the second machine learning model, a comprehensive evaluation of the model performance is performed, and the final extrusion deformation level prediction model (i.e., the tunnel extrusion deformation prediction model) is obtained based on the evaluation results.

[0094] Figure 5 A structural block diagram of a device for constructing a tunnel extrusion deformation prediction model provided in an embodiment of this disclosure. See also... Figure 5 ,include:

[0095] The acquisition module 21 is used to acquire case data on tunnel extrusion deformation and generate an initial dataset.

[0096] The data filling module 22 is used to fill the initial dataset with data using multiple data filling methods to generate multiple complete datasets.

[0097] The model training module 23 is used to train multiple first machine learning models based on the complete dataset and determine the model training dataset based on the model performance indicators of the multiple first machine learning models; to train multiple second machine learning models based on the training set in the model training dataset, and to optimize the hyperparameters in the training process of the second machine learning models using the improved Blackwing Kite algorithm; and to determine the tunnel extrusion deformation prediction model based on the model performance indicators of the multiple second machine learning models obtained from the training.

[0098] Optionally, the data population module 22 is used for:

[0099] The initial dataset was filled using the mean imputation method, median imputation method, mode imputation method, multiple imputation method, and KNN imputation method respectively, resulting in multiple complete datasets.

[0100] Optionally, the model training module 23 is used for:

[0101] For each complete dataset, multiple machine learning models are trained using different machine learning algorithms;

[0102] Calculate the sum of model performance metrics for the machine learning model, including ACC, F1, Kappa, and MCC.

[0103] The machine learning model with the largest sum of its performance metrics is identified, and the complete dataset used for the model training dataset is the machine learning model with the largest sum of its performance metrics.

[0104] Optionally, the model training module 23 is used for:

[0105] Using F1 as the iterative optimization objective, an improved black-winged kite algorithm is employed to optimize the hyperparameters during the training process of multiple machine learning models. The improved black-winged kite algorithm is obtained by using a logistic mapping to improve the initial population position of the black-winged kite algorithm. The initial population position of the improved black-winged kite algorithm is... ;

[0106] Based on the training set in the model training dataset, multiple second machine learning models are obtained by using the ten-fold cross-validation training method and multiple optimized machine learning model training algorithms.

[0107] Optionally, the model training module 23 is used for:

[0108] Based on the second machine learning model, the model performance metrics of the training set of the second machine learning model are obtained, including ACC, F1, Kappa, and MCC.

[0109] The second machine learning model is tested based on the test set to obtain the model performance metrics of the second machine learning model on the test set.

[0110] The model performance metrics of the second machine learning model are summed from the training and testing sets to obtain the sum of the model performance metrics of the second machine learning model. The second machine learning model with the largest sum of model performance metrics is the tunnel extrusion deformation prediction model.

[0111] Figure 6 This is a structural block diagram of an electronic device provided in an embodiment of this disclosure. See also... Figure 6 Electronic devices may include Figure 5 The device for constructing the tunnel compression deformation prediction model. Typically, the electronic device includes a processor 31 and a memory 32.

[0112] Processor 31 may include one or more processing cores, such as a 4-core processor or an 8-core processor. Processor 31 may be implemented using at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), or PLA (Programmable Logic Array). Processor 31 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state.

[0113] The memory 32 may include one or more computer-readable storage media, which may be non-transitory. The memory 32 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory 32 is used to store at least one instruction, which is executed by the processor 31 to implement the method for constructing a tunnel extrusion deformation prediction model executed by an electronic device as provided in the method embodiments of this application.

[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a tunnel compression deformation prediction model, characterized in that, include: The initial dataset is generated by acquiring case data on tunnel compression deformation. The initial dataset consists of tunnel depth H, tunnel diameter D, support stiffness K, rock mass quality index Q, strength-stress ratio SSR, tunnel compression deformation level Rank, and missing feature value Null in the project. Multiple data filling methods are used to fill the initial dataset to generate multiple complete datasets; Train multiple first-order machine learning models based on a complete dataset, and determine the model training dataset based on the model performance metrics of the multiple first-order machine learning models; specifically, train multiple machine learning models using different machine learning algorithms for each complete dataset. Calculate the sum of model performance metrics for the machine learning model, wherein the model performance metrics include any one of the following: weighted average of accuracy and recall, accuracy, Matthews correlation coefficient, and Kappa coefficient; The machine learning model with the largest sum of its performance metrics is identified, and the complete dataset used for the model training dataset is the machine learning model with the largest sum of its performance metrics. Multiple second machine learning models are trained based on the model training dataset. During the model training process, an improved Black-winged Kite algorithm is used to optimize the hyperparameters of the second machine learning models. Specifically, this includes using a weighted average of precision and recall as the iterative optimization objective, and employing the improved Black-winged Kite algorithm to optimize the hyperparameters of the multiple machine learning model training algorithms. The improved Black-winged Kite algorithm is obtained by using Logistic mapping to improve the initial population position of the Black-winged Kite algorithm. The initial population position of the improved Black-winged Kite algorithm is... ; Based on the training set in the model training dataset, multiple second machine learning models are obtained by using the ten-fold cross-validation training method and multiple optimized machine learning model training algorithms. The tunnel extrusion deformation prediction model is determined based on the model performance metrics of multiple second machine learning models obtained through training; specifically, this includes: obtaining model performance metrics of the training set of the second machine learning models, wherein the model performance metrics include any one of the following: a weighted average of accuracy and recall, accuracy, Matthews correlation coefficient, and Kappa coefficient; The second machine learning model is tested based on the test set to obtain the model performance metrics of the second machine learning model on the test set. The model performance metrics of the second machine learning model are summed from the training and testing sets to obtain the sum of the model performance metrics of the second machine learning model. The second machine learning model with the largest sum of model performance metrics is the tunnel extrusion deformation prediction model.

2. The method for constructing the tunnel compression deformation prediction model according to claim 1, characterized in that, Multiple data imputation methods were used to imput the initial dataset, generating multiple complete datasets, including: The initial dataset was filled using the mean imputation method, median imputation method, mode imputation method, multiple imputation method, and KNN imputation method respectively, resulting in multiple complete datasets.

3. A device for constructing a tunnel compression deformation prediction model, characterized in that, include: The acquisition module is used to acquire case data on tunnel compression deformation and generate an initial dataset. The initial dataset consists of tunnel burial depth H, tunnel diameter D, support stiffness K, rock mass quality index Q, strength-stress ratio SSR, tunnel compression deformation level Rank, and missing feature value Null in the project. The data filling module is used to fill the initial dataset with data using multiple data filling methods to generate multiple complete datasets. The model training module is used to train multiple first machine learning models based on the complete dataset, and to determine the model training dataset based on the model performance metrics of the multiple first machine learning models; specifically, it includes: training multiple machine learning models using different machine learning algorithms for each complete dataset; Calculate the sum of model performance metrics for the machine learning model, wherein the model performance metrics include any one of the following: weighted average of accuracy and recall, accuracy, Matthews correlation coefficient, and Kappa coefficient; The machine learning model with the largest sum of its performance metrics is identified, and the complete dataset used for the model training dataset is the machine learning model with the largest sum of its performance metrics. Multiple second machine learning models are trained using the training set in the model training dataset. During the training process, an improved Black-winged Kite algorithm is used to optimize the hyperparameters of the second machine learning models. Specifically, this includes using a weighted average of precision and recall as the iterative optimization objective, and employing the improved Black-winged Kite algorithm to optimize the hyperparameters of the multiple machine learning model training algorithms. The improved Black-winged Kite algorithm is obtained by using Logistic mapping to improve the initial population position of the Black-winged Kite algorithm. The initial population position of the improved Black-winged Kite algorithm is... ; Based on the training set in the model training dataset, multiple second machine learning models are obtained by using the ten-fold cross-validation training method and multiple optimized machine learning model training algorithms. The tunnel extrusion deformation prediction model is determined based on the model performance metrics of multiple second machine learning models obtained through training. Specifically, this includes: obtaining model performance metrics of the training set of the second machine learning models, wherein the model performance metrics include any one of the following: a weighted average of accuracy and recall, accuracy, Matthews correlation coefficient, and Kappa coefficient. The second machine learning model is tested based on the test set to obtain the model performance metrics of the second machine learning model on the test set. The model performance metrics of the second machine learning model are summed from the training and testing sets to obtain the sum of the model performance metrics of the second machine learning model. The second machine learning model with the largest sum of model performance metrics is the tunnel extrusion deformation prediction model.

4. An electronic device, characterized in that, The apparatus includes the device for constructing the tunnel compression deformation prediction model as described in claim 3.

5. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one piece of program code, which is executed by a processor to implement the method for constructing a tunnel extrusion deformation prediction model as described in any one of claims 1 or 2.

Images