Typhoon rainstorm debris flow susceptibility evaluation method based on ensemble learning algorithm

Abstract

The invention provides a typhoon rainstorm debris flow susceptibility evaluation method based on an ensemble learning algorithm, and the method comprises the steps: selecting an impact factor for a research region, building an evaluation index system, and obtaining related data according to an index demand; then extracting factor data by using ArcGIS and preprocessing the factor data, screening irrelevant characteristic factors by using a factor correlation analysis method, quantitatively measuring the colinearity among multivariate factors by using a multi-colinearity analysis method and eliminating the correlation among the factors by using a KPCA (Kernel Principal Component Analysis) method; then performing oversampling on debris flow occurrence samples by using an SMOTE algorithm; then optimizing parameters of the CatBoost regression model through an integrated optimization algorithm based on WOA, ISO, CSA and IGA optimization models; k-fold cross validation is carried out through Skleen to evaluate the stability of the model, and an optimal model is identified; and finally, drawing a debris flow susceptibility map by adopting the optimal model, and analyzing the influence of rainfall data in different time ranges on debris flow susceptibility prediction accuracy.

Description

Technical Field

[0001] The invention belongs to the technical field of debris flow assessment, and in particular relates to a method for assessing the susceptibility of typhoon and rainstorm type debris flow based on an integrated learning algorithm. Background Art

[0002] Debris flow is characterized by suddenness, fast flow, and strong destructive power. It often occurs in mountainous or hilly areas with poor geology. It is a complex geological disaster caused by landslides and heavy rainfall. Its occurrence process is often closely related to factors such as heavy rainfall, topography, soil conditions, and vegetation coverage. Especially during typhoons and heavy rains, a large amount of precipitation is concentrated in a short period of time, causing the hydrological conditions of the mountains to change dramatically, thereby significantly increasing the risk of debris flow. Therefore, it is crucial to assess the susceptibility of debris flow so that preventive measures can be taken in a timely manner to reduce disaster losses.

[0003] The short-duration heavy rainfall brought by typhoons and rainstorms is a key factor in the formation of typhoon-rainstorm-type debris flows. Therefore, the selection of the time range of rainfall data has a significant impact on the accuracy of debris flow susceptibility maps. With the development of machine learning, the accuracy of debris flow risk assessment has been significantly improved. Catboost, as an efficient classification and regression tool, can handle categorical features well. However, due to the low frequency of typhoon-rainstorm-type debris flows, the scarcity of samples can easily affect the model training effect. Summary of the invention

[0004] Considering that CatBoost involves multiple hyperparameters, the performance of the model can be improved by adjusting the hyperparameters. The present invention introduces the SMOTE algorithm to solve the imbalance problem in the data set, constructs an integrated optimization algorithm based on WOA, ISO, CSA, and IGA optimization models, optimizes the parameters of the CatBoost regression model, and evaluates and compares the optimization effects of the four optimization models. In addition, the influence of rainfall data at different time scales on the prediction ability of the model is introduced, aiming to identify the rainfall data time range that is most suitable for typhoon and rainstorm type debris flow prediction, improve the prediction accuracy of the typhoon and rainstorm type debris flow susceptibility map, and provide scientific support for disaster prevention and mitigation.

[0005] The evaluation method provided includes the following steps: 1) selecting influencing factors for the study area, and obtaining remote sensing images, raster data, topographic information, hydrogeological data, rainfall data of two different time lengths and historical debris flow records of the study area according to the index requirements; 2) using ArcGIS to extract factor data and preprocess, screening characteristic factors through factor correlation analysis, applying multicollinearity to perform factor multivariate collinearity analysis, and using KPCA method to eliminate correlation; 3) using SMOTE algorithm to oversample debris flow occurrence samples to balance the category distribution in the data set; 4) constructing an integrated optimization algorithm based on WOA, ISO, CSA, and IGA optimization models to optimize the parameters of the CatBoost regression model; 5) using Sklearn to evaluate model performance and identify the best model; 6) using the best model to draw a debris flow susceptibility map, and analyzing the influence of rainfall data in different time ranges on the accuracy of debris flow susceptibility prediction. The present invention improves the prediction accuracy of the typhoon and rainstorm debris flow susceptibility map by constructing an index system, balancing the data set, optimizing model parameters, and analyzing the time range of rainfall data.

[0006] The technical solution specifically adopted by the present invention to solve the technical problem is:

[0007] A method for evaluating the susceptibility of typhoon-rainstorm debris flow based on an integrated learning algorithm: Influencing factors are selected for the study area, an evaluation index system is established, and remote sensing images, raster data, topographic information, hydrogeological data, rainfall data of two different time lengths, and historical debris flow disaster records of the study area are obtained according to the index requirements; then ArcGIS is used to extract and preprocess the factor data, and the factor correlation analysis method is used to screen the uncorrelated characteristic factors, and the multicollinearity analysis method is used to quantitatively measure the collinearity between multivariate factors, and the KPCA method is used to eliminate the correlation between factors; then the SMOTE algorithm is used to oversample the debris flow occurrence samples to balance the category distribution in the data set; then the CatBoost regression model parameters are optimized through an integrated optimization algorithm based on WOA, ISO, CSA, and IGA optimization models; and Sklearn is used to perform k-fold cross validation to evaluate the model stability, calculate the accuracy, mean variance, and AUC value to evaluate the model performance, and identify the best model; finally, the best model is used to draw a debris flow susceptibility map, and the influence of rainfall data in different time ranges on the accuracy of debris flow susceptibility prediction is analyzed.

[0008] Furthermore, the influencing factors are selected for the study area, an evaluation index system is established, and remote sensing images, raster data, topographic and geomorphic information, hydrogeological data, rainfall data of two different time lengths, and historical debris flow disaster records of the study area are obtained according to the index requirements. Specifically, debris flow influencing factors are extracted for the study area at multiple scales, and the factors affecting the occurrence of debris flows are classified and integrated into topographic and geomorphic factors, geological structural factors, hydrological factors, human factors, and induced factors to establish an evaluation index system that conforms to the characteristics of typhoon and rainstorm type debris flows; remote sensing images, raster data, topographic and geomorphic information, hydrogeological data, rainfall data of two different time lengths, and historical debris flow disaster records of the study area are obtained according to the factors.

[0009] Furthermore, the method uses ArcGIS to extract factor data and perform preprocessing, screens out uncorrelated characteristic factors through factor correlation analysis, quantitatively measures the collinearity between multivariate factors through multicollinearity analysis, and eliminates the correlation between factors through KPCA method. Specifically, ArcGIS is used to extract factor data and perform preprocessing, including data cleaning, processing missing values, outliers, and normalization; Spearman rank correlation coefficient analysis is used to screen out uncorrelated characteristic factors, and variance inflation factor VIF and tolerance TOL of multicollinearity analysis method are used to perform factor multivariate collinearity analysis, and kernel principal component analysis KPCA method is used to eliminate the correlation between factors.

[0010] Furthermore, the use of the SMOTE algorithm to oversample the debris flow occurrence samples to balance the category distribution in the data set is specifically: using the synthetic minority oversampling technology SMOTE to oversample the debris flow occurrence samples, and generating new synthetic samples by interpolation in the feature space of the minority class samples, thereby increasing the number of minority class samples, making the category distribution in the data set more balanced.

[0011] Furthermore, the optimization of the parameters of the CatBoost regression model by an integrated optimization algorithm based on WOA, ISO, CSA, and IGA optimization models is specifically as follows: first, an integrated algorithm based on the whale optimization algorithm WOA, the improved snake optimization algorithm ISO, the cuckoo search algorithm CSA, and the improved genetic algorithm IGA optimization model is constructed to optimize the six hyperparameters of the Catboost regression model, namely, the maximum number of decision trees iterations, the total learning_rate affecting training time, the depth of the tree depth, the regular sub-parameter 12_leaf_reg, the number of numerical feature segmentations border_count, and the number of category feature segmentations ctr_border_count, and set initial parameters for each optimization model, including population size, number of iterations, convergence conditions, etc., and iteratively search for the optimal parameter combination in turn.

[0012] Furthermore, the k-fold cross validation is performed by Sklearn to evaluate the model stability, and the accuracy, mean variance, and AUC value are calculated to evaluate the model performance. In the process of identifying the best model: the model prediction performance is judged by calculating the accuracy and mean variance; and the ROC curve is drawn and the area under the ROC curve is calculated as the AUC value to evaluate the comprehensive performance of the model.

[0013] Furthermore, the method of using the optimal model to draw a debris flow susceptibility map and analyzing the influence of rainfall data in different time ranges on the accuracy of debris flow susceptibility prediction is specifically as follows: using the optimal model to draw a debris flow susceptibility map, analyzing the influence of rainfall data in different time ranges on the accuracy of debris flow susceptibility prediction, and using frequency ratio analysis to determine the optimal time range for using rainfall data for typhoon and rainstorm type debris flow, so as to generate a debris flow susceptibility map.

[0014] And, a typhoon-rainstorm debris flow susceptibility evaluation system based on an integrated learning algorithm, including: a data acquisition module, which is used to select influencing factors for the study area, establish an evaluation index system, and obtain remote sensing images, raster data, topographic information, hydrogeological data, rainfall data of two different time lengths, and historical debris flow disaster records of the study area according to the index requirements; a factor processing module, which is used to extract factor data using ArcGIS and perform preprocessing, screen out uncorrelated characteristic factors through factor correlation analysis methods, apply multicollinearity analysis methods to quantitatively measure the collinearity between multivariate factors, and use KPCA method to eliminate the correlation between factors. The oversampling module uses the SMOTE algorithm to oversample debris flow samples to balance the category distribution in the data set; the integrated optimization module is used to optimize the parameters of the CatBoost regression model through an integrated optimization algorithm based on the WOA, ISO, CSA, and IGA optimization models; the best model identification module is used to evaluate the model stability through k-fold cross validation using Sklearn, calculate the accuracy, average variance, and AUC value to evaluate the model performance, and identify the best model; the evaluation module uses the best model to draw a debris flow susceptibility map and analyze the impact of rainfall data in different time ranges on the accuracy of debris flow susceptibility prediction.

[0015] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of a method for evaluating the susceptibility of typhoon and rainstorm type debris flow based on an integrated learning algorithm as described above are implemented.

[0016] A non-transitory computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for evaluating the susceptibility of typhoon and rainstorm type debris flow based on an integrated learning algorithm as described above.

[0017] Compared with the prior art, the present invention and its preferred solution aim at the characteristics of low frequency, short duration and heavy rainfall of typhoon-rainstorm-type debris flow, which leads to the problem of data scarcity and uncertainty in selecting evaluation influencing factors, and proposes a typhoon-rainstorm-type debris flow susceptibility evaluation method based on an integrated optimization algorithm. The data set is enhanced by using the SMOTE algorithm, and an integrated optimization algorithm based on WOA, ISO, CSA, and IGA optimization models is constructed to optimize the parameters of the CatBoost regression model to improve the classification accuracy, generalization ability and performance stability of the model. At the same time, the present invention also analyzes the effect of rainfall data of different time scales on the prediction accuracy of the model, determines the rainfall data time scale that is most suitable for typhoon-rainstorm-type debris flow prediction, and establishes an evaluation index system for typhoon-rainstorm-type debris flow. Further improve the prediction accuracy of the susceptibility map, and provide a scientific decision-making basis for disaster prevention and mitigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0019] Figure 1 The figure is a flow chart of a method according to an embodiment of the present invention.

[0020] Figure 2 This is a diagram of the sample expansion generation process of the SMOTE algorithm according to an embodiment of the present invention.

[0021] Figure 3 This is a main process diagram of the WOA optimization algorithm according to an embodiment of the present invention.

[0022] Figure 4 This is a main process diagram of the ISO optimization algorithm according to an embodiment of the present invention.

[0023] Figure 5 This is a main process diagram of the CSA optimization algorithm according to an embodiment of the present invention.

[0024] Figure 6 This is a main process diagram of the IGA optimization algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to make the features and advantages of this patent more obvious and easy to understand, the following embodiments are specifically described in detail as follows:

[0026] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0027] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0028] In view of the characteristics of typhoon-rainstorm-type debris flows, such as low frequency, short duration, and heavy rainfall, which lead to data scarcity and uncertainty in the selection of evaluation influencing factors, the present invention proposes a typhoon-rainstorm-type debris flow susceptibility evaluation method based on an integrated optimization algorithm. The data set is enhanced by using the SMOTE algorithm, and an integrated optimization algorithm based on the WOA, ISO, CSA, and IGA optimization models is constructed to optimize the parameters of the CatBoost regression model. At the same time, the present invention also analyzes the effect of rainfall data at different time scales on the accuracy of model prediction, determines the rainfall data time scale that is most suitable for typhoon-rainstorm-type debris flow prediction, and establishes an evaluation index system for typhoon-rainstorm-type debris flow. The prediction accuracy of the susceptibility map is further improved, providing a scientific decision-making basis for disaster prevention and mitigation. The specific implementation flow chart is as follows. Figure 1 As shown. It mainly includes the following 6 steps:

[0029] Step 1: Select debris flow influencing factors for the study area, establish an evaluation index system, and obtain remote sensing images, raster data, topographic information, hydrogeological data, rainfall data of two different time lengths, and historical debris flow disaster records of the study area according to the index requirements;

[0030] Step 2: ArcGIS was used to extract factor data and preprocess them, and the factor correlation analysis method was used to screen out the characteristic factors with no correlation. The multicollinearity analysis method was used to quantitatively measure the collinearity between the multivariate factors and the KPCA method was used to eliminate the correlation between the factors.

[0031] Step 3: Use the SMOTE algorithm to oversample debris flow occurrence samples to balance the category distribution in the data set and lay the foundation for model training;

[0032] Step 4: Construct an integrated optimization algorithm based on WOA, ISO, CSA, and IGA optimization models to optimize the parameters of the CatBoost regression model;

[0033] Step 5: Use Sklearn to perform k-fold cross validation to evaluate the model stability, and calculate the accuracy, mean variance, and AUC values ​​to evaluate the model performance and identify the best model;

[0034] Step 6: Use the best model to draw a debris flow susceptibility map, analyze the impact of rainfall data in different time ranges on the accuracy of debris flow susceptibility prediction, and determine the best time range for using rainfall data to study typhoon and rainstorm debris flows.

[0035] Select a prefecture-level city in the southeast coast and give a detailed introduction according to the method flow:

[0036] Step 1: Extract debris flow influencing factors at multiple scales in the study area and classify and integrate the factors affecting the occurrence of debris flow into topographic factors, geological structural factors, hydrological factors, human factors, and induced factors, and establish an evaluation index system that conforms to the characteristics of typhoon and rainstorm debris flow. Among them, topographic factors include elevation, slope, slope aspect, curvature, NDVI, and FVC; geological structural factors include soil type, stratum lithology, water system density, distance from fault zone, and water system distance; hydrological factors include SPI and TWI; human factors include land use classification, road density, and distance from road; induced factors include annual rainfall and rainfall from June to October. According to the factors, remote sensing images, raster data, topographic information, hydrogeological data, rainfall data of two different time lengths, and historical debris flow disaster records of the study area are obtained.

[0037] Step 2: ArcGIS was used to extract factor data and perform preprocessing, including data cleaning, processing missing values, outliers, and normalization. The factor correlation analysis method, namely the Spearman rank correlation coefficient analysis, was used to screen out the characteristic factors without correlation. When the Spearman rank correlation coefficient was greater than 0.5 or less than -0.5, it was considered that there was correlation between the indicators. The multicollinearity analysis method, namely the variance inflation factor (VIF) and tolerance (TOL), was used to perform factor multicollinearity analysis. When VIF was greater than or equal to 5 or TOL was less than or equal to 0.2, it was considered that there was collinearity between the indicators, otherwise it did not exist. The KPCA method was used to eliminate the correlation between the factors to ensure that the selected factors were representative. The final indicator evaluation system includes: elevation, slope, aspect, curvature, NDVI, stratum lithology, water system density, distance from the fault zone, SPI, TWI, land use classification, road density, annual rainfall or rainfall from June to October.

[0038] S2.1: ArcGIS software was used to extract the geospatial data of the study area, including remote sensing images, raster data, topographic information, and hydrogeological data. The extracted data was then cleaned, including removing duplicate records and inconsistent data entries, and dealing with missing values, i.e., using interpolation methods or deleting records with missing values. Finally, the data was normalized to make data of different dimensions comparable, using the following formula:

[0039]

[0040] Where X is the original data and X' is the normalized data

[0041] S2.2: The Spearman rank correlation coefficient analysis method is used to measure the correlation between characteristic factors and screen out characteristic factors without correlation. The formula is:

[0042]

[0043] where ρ is the Spearman rank correlation coefficient; R i and S i are the levels of the values ​​of the observation value i; and are the average ranks of variables x and y, respectively; d i =R i -S i is the number of rank differences between two columns of paired variables; n is the total number of observations.

[0044] S2.3: Variance inflation factor (VIF) and tolerance (TOL) are used to perform factor multivariate collinearity analysis. The formula is:

[0045]

[0046] Where R 2 is the sample determination coefficient; is the predicted value; y i is the true value.

[0047] S2.4: Use the kernel principal component analysis (KPCA) method to eliminate the correlation between factors. KPCA finds nonlinear relationships by mapping data into a high-dimensional space to ensure that the selected factors are representative. The calculation process is:

[0048] ⑴ Calculate the matrix K through the kernel function, whose element K ij Defined as K ij = k(x i ,x j ). where x i and x j is the sample of the original space, k(·,·) is the kernel function.

[0049] ⑵ Calculate the eigenvalues ​​of matrix K and arrange them from large to small. Find the eigenvector α corresponding to the eigenvalue (l) (represents the lth eigenvector), and for α (l) Normalize ||α (l) ||=1.

[0050] ⑶The coordinates of the original sample under the lth nonlinear principal component are:

[0051]

[0052] Step 3: Use synthetic minority oversampling technique (SMOTE) to oversample debris flow occurrence samples, such as Figure 2 As shown, new synthetic samples are generated by interpolation in the feature space of minority class samples, thereby increasing the number of minority class samples, making the category distribution in the data set more balanced, and laying the foundation for model training. It includes the following steps:

[0053] S3.1: For each sample x in the minority class, calculate its distance to all samples in the minority class sample set using the Euclidean distance as the criterion to obtain its k nearest neighbors.

[0054] S3.2: Set a sampling ratio according to the sample imbalance ratio to determine the sampling rate N. For each minority class sample x, randomly select several samples from its k nearest neighbors. Assume that the selected nearest neighbor is o.

[0055] S3.3: For each randomly selected neighbor o, construct a new sample with the original sample according to the following formula:

[0056] o(new)=o+rand(0,1)×(xo) (6)

[0057] Step 4: Construct an integrated algorithm based on the whale optimization algorithm (WOA), improved snake optimization algorithm (ISO), cuckoo search algorithm (CSA), and improved genetic algorithm (IGA) optimization models, and optimize the six hyperparameters of the Catboost regression model, including the maximum number of decision trees iterations, the total training time learning_rate, the depth of the tree depth, the regular sub-parameter 12_leaf_reg, the number of numerical feature segmentations border_count, and the number of category feature segmentations ctr_border_count. Set the initial parameters for each optimization model, including population size, number of iterations, convergence conditions, etc., and iterate and search for the optimal parameter combination in turn. The population size of the whale optimization algorithm (WOA) is 50 and the dimension is 2; the population size of the improved snake optimization algorithm (ISO) is set to 50 and the number of iterations is 1000; the levy distribution parameter of the cuckoo search algorithm (CSA) is 1.5, the step size is 0.01, the number of nests is 20, and the maximum number of iterations is 500; the initial population size, crossover probability and mutation probability of the improved genetic algorithm (IGA) are configured to 50, 0.8 and 0.05 respectively. The algorithm flow is as follows Figure 3-Figure 6 shown.

[0058] The four algorithm formulas are as follows:

[0059] ① The Whale Optimization Algorithm (WOA) simulates the bubble net predation behavior and group cooperation strategy of humpback whales, performs global search and local exploration in the parameter space, and iteratively updates the potential optimal solution. It contains the following formula:

[0060]

[0061]

[0062] Where t is the current iteration number; is the position of the best candidate solution before t generations; and denote the solutions of generation t and the next generation respectively; and is the coefficient vector; and Represents two random numbers between 0 and 1; is a number that decreases linearly from 2 to 0 with each iteration; G is the total number of iterations.

[0063]

[0064] l=(a 2 -1)·r 3 +1 (15)

[0065]

[0066] Where b is a constant, usually set to 1; l is a random number between [-1,1]; r 3 Represents two random numbers between 0 and 1.

[0067]

[0068] in is a randomly selected solution; when |A|<1, the prey encirclement behavior is adopted; when |A|≥1, the random search behavior is adopted.

[0069] ② Improved Snake Optimization Algorithm (ISO) The improved snake optimization algorithm simulates the predation behavior and dynamic adaptation ability of snakes to perform global search and local exploration in the parameter space to iteratively update the potential optimal solution. It contains the following formula:

[0070] X i =X min +r*(X max -X min ) (19)

[0071] Where X i is the position of the i-th snake; r is a random number in the range [0,1]; X max , Xmin are the upper and lower bounds of the problem to be solved.

[0072]

[0073]

[0074] Where N is the size of the snake population; N m is the number of males; N f is the number of females; t is the current number of iterations; T is the maximum number of iterations; C 1 is a constant, take 0.5.

[0075] X i,m (t+1)=X rand,m (t)±C 2 *A m *((A max -A min )*rand+X min ) (twenty three)

[0076]

[0077] Where X i,m X is the male position; rand,m is the position of the randomly selected male; rand is a random number in the range [0,1]; A m The ability to find food for males; f rand,m is the position X of the randomly selected male rand,m The fitness value of i,m For male position X i,m The fitness value of 2 is a constant, taking 0.05; X i,f X is the female position; rand,f is the position of a randomly selected female; A f The ability to find food for females; f rand,f is the position X of a randomly selected female rand,f The fitness value of i,f For male position X i,f The fitness value of .

[0078] X i,j (t+1)=X food ±C 3 *Temp*rand*(X food -X i,j (t)) (26)

[0079] Where X i,j The position of the snake (male or female); X foodThe best position for snakes; C 3 is a constant, taking 0.05.

[0080] X i,m (t+1)=X i,m +C 3 *FM*rand*(Q*X best,f -X i,m (t)) (27)

[0081] X i,f (t+1)=X i,f +C 3 *FF*rand*(Q*X best,m -X i,f (t)) (28)

[0082]

[0083] X i,m (t+1)=X i,m +C 3 *M m *rand*(Q*X best,f -X i,m (t)) (31)

[0084] X i,f (t+1)=X i,f +C 3 *M f *rand*(Q*X best,m -X i,f (t)) (32)

[0085]

[0086] X worst,m =X min +rand*(X max -X min ) (35)

[0087] X worst,f =X min +rand*(X max -X min ) (36)

[0088] Where X i,m is the ith male position; X best,f is the best position of the female snake group; FM is the male fighting ability; X i,f is the ith female position; X best,m is the best position of the male snake group; FF is the fighting ability of the female; f best,f is the best position X for female snakesbest,f The fitness value of best,m is the best position X in male snake best,m The fitness value f i is the fitness value of the snake; M m 、M f X is the mating ability of males and females respectively; worst,m , X worst,f The worst positions for males and females, respectively.

[0089] ③ The Cuckoo Search Algorithm (CSA) effectively optimizes the parameters of the SVM model by simulating the parasitic reproduction behavior of cuckoos and combining global exploration with local search using Levy flight. It contains the following formulas:

[0090]

[0091] Where α is the step size scaling factor; Levy(β) is the Levy random path; is the * operation; r and ε are random numbers that follow a uniform distribution; Heaviside(x) is a jump function (when the input x is less than 0, the value of the Heaviside function is 0; when the input x is greater than or equal to 0, the value of the Heaviside function is 1); X i , X j is any other continuous bird nest; β is usually taken as 1.5.

[0092] ④The improved genetic algorithm (IGA) simulates the selection, crossover and mutation operations in the natural genetic process, combines the fitness function of the specific problem, and iteratively searches to optimize the parameters of the SVM model, thereby improving the prediction performance of the model. It contains the following formula:

[0093]

[0094] f(x)=N·[A(x)] -ln[A(x)] (42)

[0095] A(x)=f+T·P w (43)

[0096] x new =(x i1 ,x i2 ,...,x ik ,x j(k+1) ,...,x jn ) (44)

[0097] x' nk =x nk +δ,withprobability μ (45)

[0098] Where f(x i ) is the fitness function of the ith individual; N is the total number of individuals in the population; x is an individual in the population; A(x) is the evaluation function of the individual; T represents the infeasible solution of x; P w is the penalty weight of the infeasible path; δ is a random small change; μ is the mutation rate.

[0099] Step 5: Use sklearn to perform k-fold cross validation to evaluate the model stability, and calculate the accuracy, mean square error (MSE), and AUC values ​​to evaluate the model performance. After comparison, the WOA-Catboost integrated optimization algorithm performs best.

[0100] ① Calculate the accuracy and mean square error (MSE) indicators to judge the model prediction performance. The indicator formula is as follows:

[0101]

[0102] ② Draw the ROC curve and calculate the area under the ROC curve (AUC value) to evaluate the comprehensive performance of the model. The calculation formula of the curve abscissa (FPR) and ordinate (TPR) is as follows:

[0103]

[0104] Step 6: Use the WOA-Catboost integrated optimization algorithm to draw a typhoon-rainstorm type debris flow susceptibility map, analyze the impact of rainfall data in different time ranges on the accuracy of debris flow susceptibility prediction, use frequency ratio analysis to determine the optimal time range for using rainfall data for typhoon-rainstorm type debris flow, and generate a higher-precision debris flow susceptibility map.

[0105] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0106] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0107] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0108] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0109] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0110] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

[0111] This patent is not limited to the above-mentioned optimal implementation mode. Anyone can derive various other forms of a typhoon and rainstorm type debris flow susceptibility evaluation method based on an integrated learning algorithm under the inspiration of this patent. All equal changes and modifications made according to the scope of the patent application of the present invention should be covered by this patent.

Claims

1. A method for evaluating the susceptibility of typhoon and rainstorm debris flows based on an ensemble learning algorithm, characterized in that: Influencing factors were selected for the study area, and an evaluation index system was established. Based on the index requirements, remote sensing images, raster data, topographic information, hydrogeological data, rainfall data of two different time lengths, and historical debris flow disaster records of the study area were obtained. Then, ArcGIS was used to extract factor data and preprocess them. The factor correlation analysis method was used to screen out characteristic factors with no correlation. The multicollinearity analysis method was used to quantitatively measure the collinearity between multivariate factors, and the KPCA method was used to eliminate the correlation between factors. Then, the SMOTE algorithm is used to oversample the debris flow samples to balance the category distribution in the data set; the parameters of the CatBoost regression model are optimized through an integrated optimization algorithm based on the WOA, ISO, CSA, and IGA optimization models; Sklearn was used to perform k-fold cross validation to evaluate the model stability, calculate the accuracy, mean variance, and AUC value to evaluate the model performance, and identify the best model. Finally, the best model was used to draw a debris flow susceptibility map, and analyze the impact of rainfall data in different time ranges on the accuracy of debris flow susceptibility prediction.

2. The method for evaluating the susceptibility of typhoon and rainstorm debris flow based on an ensemble learning algorithm according to claim 1 is characterized in that: The method comprises the following steps: selecting influencing factors for the study area, establishing an evaluation index system, and acquiring remote sensing images, raster data, topographic and geomorphic information, hydrogeological data, rainfall data of two different time lengths, and historical debris flow disaster records of the study area according to the index requirements. Specifically, the debris flow influencing factors are extracted at multiple scales in the study area, and the factors affecting the occurrence of debris flows are classified and integrated into topographic and geomorphic factors, geological structural factors, hydrological factors, human factors, and induced factors to establish an evaluation index system that conforms to the characteristics of typhoon and rainstorm type debris flows; acquiring remote sensing images, raster data, topographic and geomorphic information, hydrogeological data, rainfall data of two different time lengths, and historical debris flow disaster records of the study area according to the factors.

3. The method for evaluating the susceptibility of typhoon and rainstorm debris flow based on an ensemble learning algorithm according to claim 1, characterized in that: The method uses ArcGIS to extract factor data and perform preprocessing, uses factor correlation analysis to screen out uncorrelated characteristic factors, uses multicollinearity analysis to quantitatively measure the collinearity between multivariate factors, and uses KPCA to eliminate the correlation between factors. Specifically, ArcGIS is used to extract factor data and perform preprocessing, including data cleaning, processing missing values, outliers, and normalization; Spearman rank correlation coefficient analysis is used to screen out uncorrelated characteristic factors, and the variance inflation factor VIF and tolerance TOL of the multicollinearity analysis method are used to perform factor multivariate collinearity analysis, and the kernel principal component analysis KPCA method is used to eliminate the correlation between factors.

4. The method for evaluating the susceptibility of typhoon and rainstorm debris flow based on an ensemble learning algorithm according to claim 1, characterized in that: The method of using the SMOTE algorithm to oversample the debris flow occurrence samples to balance the category distribution in the data set is specifically as follows: using the synthetic minority oversampling technology SMOTE to oversample the debris flow occurrence samples, and generating new synthetic samples by interpolating in the feature space of the minority class samples, thereby increasing the number of minority class samples, making the category distribution in the data set more balanced.

5. The method for evaluating the susceptibility of typhoon and rainstorm debris flow based on an ensemble learning algorithm according to claim 1, characterized in that: The method of optimizing the parameters of the CatBoost regression model by using an integrated optimization algorithm based on WOA, ISO, CSA, and IGA optimization models is as follows: first, an integrated algorithm based on the whale optimization algorithm WOA, the improved snake optimization algorithm ISO, the cuckoo search algorithm CSA, and the improved genetic algorithm IGA optimization model is constructed to optimize the six hyperparameters of the Catboost regression model, namely, the maximum number of decision trees iterations, the total learning_rate time affecting training, the depth of the tree depth, the regular sub-parameter 12_leaf_reg, the number of numerical feature segmentations border_count, and the number of category feature segmentations ctr_border_count. Initial parameters are set for each optimization model, including population size, number of iterations, convergence conditions, etc., and the optimal parameter combination is iteratively searched in turn.

6. The method for evaluating the susceptibility of typhoon and rainstorm debris flow based on an ensemble learning algorithm according to claim 1, characterized in that: The k-fold cross validation was performed by Sklearn to evaluate the model stability, and the accuracy, mean variance, and AUC value were calculated to evaluate the model performance. In the process of identifying the best model: the model prediction performance was judged by calculating the accuracy and mean variance; the ROC curve was drawn and the area under the ROC curve was calculated as the AUC value to evaluate the comprehensive performance of the model.

7. The method for evaluating the susceptibility of typhoon and rainstorm debris flow based on an ensemble learning algorithm according to claim 1, characterized in that: The method of using the best model to draw a debris flow susceptibility map and analyzing the influence of rainfall data in different time ranges on the accuracy of debris flow susceptibility prediction is specifically as follows: using the best model to draw a debris flow susceptibility map, analyzing the influence of rainfall data in different time ranges on the accuracy of debris flow susceptibility prediction, and using frequency ratio analysis to determine the best time range for using rainfall data for typhoon and rainstorm type debris flow, so as to generate a debris flow susceptibility map.

8. A typhoon and rainstorm debris flow susceptibility assessment system based on an integrated learning algorithm, characterized in that: include: The data acquisition module is used to select influencing factors for the study area, establish an evaluation index system, and obtain remote sensing images, raster data, topographic information, hydrogeological data, rainfall data of two different time lengths, and historical debris flow disaster records of the study area according to the index requirements; the factor processing module is used to extract factor data and preprocess it using ArcGIS, screen out uncorrelated characteristic factors through factor correlation analysis methods, apply multicollinearity analysis methods to quantitatively measure the collinearity between multivariate factors, and use KPCA method to eliminate the correlation between factors; The oversampling module uses the SMOTE algorithm to oversample debris flow samples to balance the category distribution in the data set; the integrated optimization module is used to optimize the parameters of the CatBoost regression model through an integrated optimization algorithm based on the WOA, ISO, CSA, and IGA optimization models; the best model identification module is used to evaluate the model stability through Sklearn k-fold cross validation, calculate the accuracy, mean variance, and AUC value to evaluate the model performance, and identify the best model; the evaluation module uses the best model to draw a debris flow susceptibility map and analyze the impact of rainfall data in different time ranges on the accuracy of debris flow susceptibility prediction.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of a method for evaluating the susceptibility of typhoon and rainstorm type debris flow based on an integrated learning algorithm as described in any one of claims 1 to 7 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for evaluating the susceptibility of typhoon and rainstorm type debris flow based on an integrated learning algorithm as described in any one of claims 1 to 7 are implemented.

Images