Coastal erosion susceptibility interpretable evaluation method based on random forest

Abstract

The invention provides a coast erosion susceptibility interpretable evaluation method based on a random forest, and the method comprises the steps: selecting disaster-pregnancy factors, and obtaining disaster-pregnancy factor data; coastline data are extracted, coast erosion areas are divided based on coastline changes, and coast erosion area distribution conditions are obtained; preprocessing the disaster-gestation factor data and the coastline data to construct a coast erosion data set, and dividing the coast erosion data set into a training set and a test set according to a proportion; constructing a coast erosion susceptibility evaluation model based on a random forest algorithm; carrying out comprehensive evaluation and comparison on the models to screen out an optimal model; based on the optimal random forest model obtained through screening, the contribution of each factor in coast erosion is analyzed and quantified through feature importance, the factors contributing to coast erosion are identified and obtained, and partial dependence and SHAP are used for carrying out interpretable analysis on the model so as to explain the influence of different disaster-gestation factors on model prediction.

Description

Technical Field

[0001] The present invention belongs to the technical fields such as geological disaster analysis and evaluation, and particularly relates to an interpretable evaluation method for coastal erosion susceptibility based on random forest. Background Art

[0002] The coastal zone is an important area where human activities interact with the natural environment, and it is also an extremely sensitive and fragile part of the ecosystem. With the intensification of climate change and human activities, coastal erosion disasters have become one of the major marine disasters faced by the global society, posing a serious threat to the safety of residents in coastal areas and restricting economic development. Therefore, evaluating the susceptibility of coastal erosion is crucial for disaster prevention and mitigation work in coastal areas. Although various models have been applied to disaster risk assessment and prediction, existing coastal erosion risk assessment models still have uncertainties, especially in expressing the non-linear relationship between driving factors and erosion riskiness and in evaluating the weight contribution of driving factors.

[0003] As an application of computer data mining technology in the field of disaster risk assessment, machine learning models can perform non-linear fitting more effectively compared to traditional mathematical statistics models, revealing the complex relationship between the occurrence of disasters and disaster-causing factors. With its ability to process large-scale data and mine complex associations, machine learning models have shown significant potential in the prediction, risk assessment, and management decision-making of coastal erosion. However, traditional machine learning models, such as random forest, support vector machine, and artificial neural network, due to their "black box" characteristics, that is, the opacity of the internal mechanism, bring difficulties and uncertainties in understanding to decision-makers. Summary of the Invention

[0004] Therefore, aiming at the defects and deficiencies of the existing technology, the present invention aims to construct a coastal erosion susceptibility model through the random forest method and screen out the optimal model from it. Through feature importance analysis, identify and quantify the influence degree of different disaster-forming factors on coastal erosion, and use an interpretable method to reveal the internal mechanism of the machine learning model, so as to achieve an accurate assessment of coastal erosion susceptibility.

[0005] The present invention provides an interpretable evaluation method for coastal erosion susceptibility based on random forest, and the method includes the following steps:

[0006] Step 1: Select disaster-forming factors and obtain data on disaster-forming factors; Step 2: Extract coastline data, divide the coastal erosion areas based on coastline changes, and obtain the distribution of coastal erosion areas; Step 3: Perform preprocessing on the data to be processed, including interpolation, resampling, unifying the resolution and coordinate system, etc., to construct a coastal erosion dataset; Step 4: Divide the obtained dataset into a training set and a test set according to a ratio, and construct a coastal erosion susceptibility evaluation model based on the random forest algorithm; Step 5: Conduct a comprehensive evaluation of the model; Step 6: Quantify the contributions of various factors to coastal erosion through feature importance analysis based on the random forest model, identify the factors that contribute greatly to the occurrence of coastal erosion, and use the partial dependence and SHAP methods to perform interpretable analysis on the model to explain the influence of different disaster-forming factors on the model prediction. This method can not only accurately predict coastal erosion, but also explain and quantify the characteristic factors that induce coastal erosion, so as to accurately evaluate the susceptibility of coastal erosion. It quantifies the contribution degrees of different disaster-forming factors to coastal erosion through the feature importance analysis method, and uses the model interpretability method to explain the internal mechanism of the machine learning black box model, and accurately evaluates the susceptibility of coastal erosion.

[0007] The technical solution specifically adopted by the present invention to solve its technical problems is as follows:

[0008] An interpretable evaluation method for coastal erosion susceptibility based on random forest: Select disaster-forming factors and obtain data on disaster-forming factors; Extract coastline data, divide the coastal erosion areas based on coastline changes, and obtain the distribution of coastal erosion areas; Preprocess the data on disaster-forming factors and coastline data to construct a coastal erosion dataset and divide it into a training set and a test set according to a ratio; Construct a coastal erosion susceptibility evaluation model based on the random forest algorithm; Conduct a comprehensive evaluation and comparison of the model to select the optimal model; Quantify the contributions of various factors to coastal erosion through feature importance analysis based on the selected optimal random forest model, identify the factors that contribute to the occurrence of coastal erosion, and use partial dependence and SHAP to perform interpretable analysis on the model to explain the influence of different disaster-forming factors on the model prediction.

[0009] Furthermore, the selected disaster-forming factors include: elevation, slope, soil properties, topography, land use type, population density, GDP, annual rainfall, sea level height, significant wave height, wave period, wave direction, suspended sediment concentration, tidal range, wind speed, and coastline type.

[0010] Further, for the extraction of coastline data and the division of coastal erosion areas based on coastline changes to obtain the distribution of coastal erosion areas, the specific steps are as follows: Based on remote sensing satellite images, use ENVI to extract the coastline data of a given area. According to the change in the shoreline position between two adjacent periods, identify the area where the shoreline has retreated landward in the later period compared to the previous period as the erosion area, and sample erosion points with a given grid size as the basic unit; identify the area where the shoreline has advanced seaward or remained unchanged in the later period compared to the previous period as the non-erosion area, and sample non-erosion points. Thus, obtain the distribution of coastal erosion areas and construct a historical list of coastal erosion in the study area.

[0011] Further, the preprocessing is specifically as follows: Interpolate and resample the obtained disaster-causing factor data, resample the resolution of all factors to a unified spatial resolution scale consistent with the basic unit, and save the vector and raster data with a unified projection coordinate. For areas with missing data, use Kriging method for interpolation and supplementation, and then perform multiple collinearity detection and data standardization processing of the disaster-causing factors; smooth the extracted coastline data.

[0012] Further, the specific method for smoothing the extracted shoreline is as follows: Use the Digital Shoreline Analysis System module on the ArcGIS platform to establish a monitoring shoreline retreat rate trend model; when calculating the change rate of the shoreline, use the end point change rate and the linear regression change rate as indicators.

[0013] Further, the specific method for constructing a coastal erosion susceptibility evaluation model based on the random forest algorithm is as follows: Conduct research and analysis on the binary classification problem of coastal erosion based on random forest: erosion / non-erosion.

[0014] Further, the specific method for comprehensively evaluating and comparing the models to select the optimal model is as follows: Based on the test data set, use accuracy, precision, recall rate, F1 score, and Kappa index to evaluate the effectiveness of the model; based on the test data set, use the Receiver Operating Characteristic curve ROC to evaluate the performance of the model.

[0015] Further, after identifying the factors contributing to coastal erosion, sort the feature importance degrees of each factor; perform interpretable analysis on the model through partial dependence and SHAP to explain and quantify the characteristic factors inducing coastal erosion.

[0016] And, an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the steps of the above-mentioned interpretable evaluation method for coastal erosion susceptibility based on random forest.

[0017] A non-transitory computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, the steps of an interpretable evaluation method for coastal erosion susceptibility based on a random forest as described above are implemented.

[0018] Compared with the prior art, the beneficial effects of the present invention and its preferred solutions at least include:

[0019] 1. By constructing a model through a random forest machine learning algorithm and conducting a comprehensive evaluation of the model, the present invention can more accurately predict the areas prone to coastal erosion, improving the accuracy and reliability of the prediction compared with traditional methods.

[0020] 2. Through feature importance analysis, the contribution degrees of different disaster-forming factors to coastal erosion are quantified, which helps to identify the key factors that have the greatest impact on the occurrence of coastal erosion and provides a scientific basis for formulating disaster prevention and mitigation measures.

[0021] 3. Two methods, partial dependence and SHAP value, are used to conduct interpretable analysis on the model, revealing the internal mechanism of the machine learning model, solving the problem that traditional "black box" models are difficult to explain, and enhancing the transparency and credibility of the model.

[0022] 4. By accurately evaluating the susceptibility of coastal erosion and explaining the influencing factors of model prediction, it provides strong support for disaster management decisions in coastal areas, and helps to formulate more effective disaster prevention and mitigation strategies and measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The present invention will be further described in detail below in conjunction with the drawings and specific embodiments:

[0024] Figure 1 is a schematic flowchart of an interpretable evaluation method for coastal erosion susceptibility based on a random forest according to an embodiment of the present invention.

[0025] Figure 2 is a feature importance diagram of disaster-forming factors for coastal erosion according to an embodiment of the present invention.

[0026] Figure 3 is a partial dependence diagram of disaster-forming factors for coastal erosion according to an embodiment of the present invention.

[0027] Figure 4 is a SHAP diagram of disaster-forming factors for coastal erosion according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] To make the features and advantages of this patent more obvious and understandable, specific embodiments are hereinafter given and described in detail as follows:

[0029] It should be noted that the following detailed description is illustrative and aims to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs.

[0030] 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 "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0031] An embodiment of the present invention provides an interpretable assessment method for coastal erosion susceptibility based on random forest, as Figure 1 shown, providing implementation solutions for the following steps:

[0032] Step 1: Select disaster-forming factors and obtain disaster-forming factor data;

[0033] In Step 1, the selected disaster-forming factors are considered from aspects such as natural geographical conditions, marine dynamic factors, climate factors, human activities, and nearshore sediment supply. Finally, 16 factors including elevation, slope, soil properties, topography and geomorphology, land use type, population density, GDP, annual rainfall, sea level height, significant wave height, wave period, wave direction, suspended sediment concentration, tidal range, wind speed, and coastline type are selected.

[0034] Step 2: Extract coastline data, divide the coastal erosion area based on coastline changes, and obtain the distribution of the coastal erosion area;

[0035] Specifically, in Step 2: According to the remote sensing satellite images of the Landsat series, the coastline data of the study area is extracted based on ENVI. According to the change in the shoreline position between two adjacent periods, the area where the shoreline retreats landward in the latter period relative to the former period is identified as the erosion area, and erosion points are sampled with a basic unit of 30m×30m; the area where the shoreline advances seaward or the shoreline does not change in the latter period relative to the former period is identified as the non-erosion area, and non-erosion points are sampled. Thus, the distribution of the coastal erosion area is obtained, and a historical list of coastal erosion in the study area is constructed.

[0036] Step 3: Perform preprocessing on the data to be processed, including interpolation, resampling, unifying the resolution and coordinate system, etc., and construct a coastal erosion dataset;

[0037] Step 3 specifically involves: uniformly processing the obtained disaster-causing factor data on the ArcGIS platform. Through interpolation and resampling, the resolution of all factors is resampled to a unified spatial resolution scale consistent with the basic unit, and vector and raster data with a unified projection coordinate (WGS_1984_UTM_Zone_50N) are saved. For areas with missing data, Kriging interpolation is used for supplementation. Subsequently, multicollinearity detection and data standardization processing of the disaster-causing factors are carried out. The extracted coastline data needs to be smoothed to meet the requirements of subsequent analysis.

[0038] In Step 3, multicollinearity refers to the situation where there is a high degree of correlation among the explanatory variables in a multiple linear regression model. The variance inflation factor (VIF) and tolerance (TOL) are widely used for multicollinearity detection, and they are reciprocals of each other. The formula for calculating the VIF of a variable is as follows:

[0039]

[0040] In formula (1), R 2 is the square of the multiple correlation coefficient of the explanatory variable, that is, the coefficient of determination obtained when this variable is used as the dependent variable and all other variables are used as independent variables for multiple linear regression.

[0041] In Step 3, data standardization processing is to eliminate the influence of dimension and the influence of the variation size and numerical size of the variable itself. The "Min-Max standardization" method is adopted, also known as normalization processing, which is a linear transformation of the original data to map all eigenvalue to the range of [0,1]. Its formula is:

[0042]

[0043] In the formula, x Max is the minimum value of the data; x Min is the maximum value of the data.

[0044] In Step 3, for the smoothing process of the extracted shoreline, a monitoring shoreline recession rate trend model is established using the Digital Shoreline Analysis System module (DSAS) on the ArcGIS platform. When using DSAS to calculate the change rate of the shoreline, the End Point Rate (EPR) and Linear Regression Rate (LRR) are used as indicators. The formula for EPR is:

[0045]

[0046] In the formula, NSM is the distance of the coastline from the baseline in the farthest year and the nearest year, m; SP is the time interval between the nearest year and the farthest year, yr.

[0047] The LRR is determined by fitting the least squares method to the points where the profile line intersects the coastline, and the change rate of the coastline is calculated. The linear regression method uses all the data without considering the changes in trends and accuracies. The calculation formula is:

[0048] y = ax + b (4)

[0049]

[0050] In the formula, a and b are the slope and intercept of the fitted line of the shoreline position sequence respectively, x i is the coastline position at the i-th period, and y i is the distance interval between the shoreline point and the baseline point at the i-th period on a certain profile perpendicular to the true coastline, and n is the number of coastline phases.

[0051] Step 4: Divide the obtained dataset into a training set and a test set according to a ratio, and construct a coastal erosion susceptibility evaluation model based on the random forest algorithm;

[0052] Specifically, in Step 4: Randomly divide the obtained dataset into a training set and a test set according to a ratio of 8:2, and conduct research and analysis on the binary classification problem (erosion / non-erosion) of coastal erosion based on the random forest (RF) model.

[0053] Step 5: Conduct a comprehensive evaluation of the random forest model, compare and screen out the optimal model;

[0054] Specifically, in Step 5: Conduct a comprehensive evaluation of the random forest model. Based on the test dataset, accuracy, precision, recall, F1-score, and Kappa index are used to evaluate the effectiveness of the model. The calculation formulas for each index are:

[0055]

[0056] In the formula, TP represents true positive; FP represents false positive; FN represents false negative; TN represents true negative. The value ranges of accuracy, precision, recall, and F1-score are [0, 1]. Accuracy is used to measure the percentage of the number of correct predictions in all the total predictions. The closer the accuracy of the model is to 1, the higher the overall accuracy of the model; the closer the precision of the model is to 1, the smaller the probability of misjudgment in the results predicted by the model as erosion; the closer the recall of the model is to 1, the stronger the prediction ability of the model for erosion; F1-score combines precision and recall, and the larger its value, the better the prediction performance of the model.

[0057] Based on the test dataset, the performance of the above model was evaluated using the Receiver Operating Characteristic (ROC) curve. The Receiver Operating Characteristic curve is a tool commonly used to evaluate the performance of binary classification models. It is a curve plotted with the True Positive Rate as the vertical axis and the False Positive Rate as the horizontal axis. The area under the ROC curve is the AUC value, and the larger the AUC value, the better the performance of the model. The optimal model was selected according to the above indicators.

[0058] Step 6: Based on the optimal model, quantify the contributions of various factors to coastal erosion through feature importance analysis, identify the factors that contribute significantly to the occurrence of coastal erosion, and use two methods, partial dependence and SHAP, to perform interpretable analysis on the model to explain the influence of different disaster-forming factors on the model prediction.

[0059] Specifically, Step 6 is as follows: Conduct feature importance analysis of the disaster-forming factors of coastal erosion based on the optimal model, quantify the contributions of different factors to the occurrence of erosion, so as to identify the factors that have a significant impact on the occurrence of coastal erosion disasters in a certain estuary at different times, and rank the degree of feature importance of each factor. Perform interpretable analysis on the model through two methods, partial dependence and SHAP, so as to know how each disaster-forming factor affects the prediction results of coastal erosion.

[0060] In Step 6, the Partial Dependence Plot (PDP) shows the marginal effect of a single feature factor on the prediction results of a machine learning model, and can clearly display the association pattern between the target variable and different feature factors, whether it is linear, monotonic or more complex non-linear relationships.

[0061] The partial dependence function is defined as:

[0062]

[0063] In the formula, x s is the feature variable for which the partial correlation function needs to be plotted, and x C are the other feature variables used in the machine learning model. The feature variables in set S are the features whose influence on the prediction is desired to be known. By integrating over x C , a function that only depends on x s is obtained, and this function is the partial dependence function, which can achieve the interpretation of a single variable. In actual operation, the Monte Carlo method is usually used to obtain the partial dependence function by calculating the average value of the training set. The specific formula is as follows, where n represents the sample size.

[0064]

[0065] In step 6, SHAP is a machine learning model interpretability method based on game theory and additive feature attribution. The contribution calculation formula for each feature is as follows:

[0066]

[0067] In the formula, φ i represents the contribution of the i-th coastal erosion disaster-forming factor, N represents the set of all disaster-forming factors, S represents a subset of the given disaster-forming factors, and f(S∪{i}) and f(S) represent the model results with or without the i-th disaster-forming factor. SHAP generates an interpretable model through the method of additive feature attribution, that is, the output model is defined as the linear sum of input variables, and the formula is:

[0068]

[0069] In the formula, z'∈{0,1} M , which is equal to 1 when the sample contains the disaster-forming factor i, and otherwise equal to 0; M is the number of input disaster-forming factors; φ 0 is the average prediction value; φ i is the contribution value of the i-th disaster-forming factor, that is, the SHAP value. A positive SHAP value indicates a positive contribution of this factor to the occurrence of erosion, and a negative value indicates a negative contribution.

[0070] The above solution of this embodiment is further introduced and demonstrated through a more specific test example as follows:

[0071] An evaluation model for the susceptibility of coastal erosion from 2010 to 2015 is established for a certain estuary, and interpretability evaluation is carried out, which specifically includes the following steps:

[0072] Step 1: According to the regional characteristics of a certain estuary, 16 factors are selected, including elevation, slope, soil properties, topography and geomorphology, land use type, population density, GDP, annual rainfall, sea level height, significant wave height, wave period, wave direction, suspended sediment concentration, tidal range, wind speed, and coastline type.

[0073] Step 2: Based on the remote sensing satellite images of the Landsat series from 2015 to 2020, the coastline data of a certain estuary is extracted based on ENVI. According to the change in the shoreline position between two adjacent periods, the area where the shoreline retreats landward in the later period compared to the previous period is identified as the erosion area, and erosion points are sampled with a basic unit of 30m×30m; the area where the shoreline advances seaward or the shoreline does not change in the later period compared to the previous period is identified as the non-erosion area, and non-erosion points are sampled. Thus, the distribution of the coastal erosion area is obtained, and a historical inventory of coastal erosion from 2015 to 2020 in a certain estuary is constructed.

[0074] Step 3: Process the above-mentioned step data uniformly on the ArcGIS platform. Through interpolation and resampling, resample the resolution of all factors to a unified spatial resolution scale consistent with the basic unit, and save the vector and raster data with a unified projection coordinate (WGS_1984_UTM_Zone_50N). For the areas with missing data, use the Kriging method for interpolation and supplementation, and then conduct multiple collinearity detection and data standardization processing of the disaster-forming factors. The extracted coastline data needs to be smoothed to meet the requirements of subsequent analysis.

[0075] Step 4: Randomly divide the obtained dataset into a training set and a test set according to a ratio of 8:2, and conduct research and analysis on the binary classification problem (erosion / non-erosion) of coastal erosion based on the Random Forest (RF) model.

[0076] Step 5: Conduct a comprehensive evaluation of the Random Forest model. Based on the test dataset, use accuracy, precision, recall rate, F1 score, and Kappa index to evaluate the effectiveness of the model. Based on the test dataset, use the Receiver Operating Characteristic curve (ROC) to evaluate the performance of the above model. The Receiver Operating Characteristic curve is a tool commonly used to evaluate the performance of binary classification models. It is a curve plotted with the True Positive Rate as the ordinate and the False Positive Rate as the abscissa. The area under the ROC curve is the AUC value. The larger the AUC value, the better the performance of the model. Select the optimal model according to the above indicators.

[0077] Table 1 shows the comparison of the evaluation metric parameters of the coastal erosion susceptibility model established based on the Random Forest. The numerical values of the specific evaluation indicators are as follows:

[0078] period model accuracy precision recall F1 score Kappa index AUC 2010-2015 RF 0.93 0.96 0.90 0.92 0.85 0.98

[0079] Table 1. Comparison of Evaluation Indicators of Different Coastal Erosion Susceptibility Models

[0080] It can be seen from Table 1 that the Random Forest model has a good classification effect, high prediction accuracy, and strong generalization ability. It can be concluded that the Random Forest model has high prediction accuracy and excellent performance in the evaluation of the susceptibility of a certain estuary coastal erosion.

[0081] Step 6: Conduct an analysis of the importance of the characteristics of the disaster-forming factors of coastal erosion based on the Random Forest model, quantify the contributions of different factors to the occurrence of erosion, so as to identify the factors that have a significant impact on the occurrence of coastal erosion disasters in a certain estuary at different times, and rank the degree of importance of the characteristics of each factor, as Figure 2 shown. Conduct an interpretable analysis of the model through two methods: partial dependence and SHAP, so as to know how each disaster-forming factor affects the prediction result of coastal erosion. As Figure 3 、4 as shown

[0082] Based on the same inventive concept, the present invention further provides a computer device, which includes: one or more processors, and a memory for storing one or more computer programs; the program includes program instructions, and the processor is configured to execute the program instructions stored in the memory. The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and is used to implement one or more instructions. Specifically, it is used to load and execute one or more instructions in the computer storage medium to implement the above method.

[0083] It should be further noted that, based on the same inventive concept, the present invention further provides a computer storage medium, on which a computer program is stored, and the computer program, when run by a processor, executes the above method. The storage medium may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electrical, magnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive list) of the computer-readable storage medium include: an electrical connection having one or more wires, a portable computer disk, a hard disk, a Random Access Memory (RAM), a Read-Only Memory (ROM), an Erasable Programmable Read-Only Memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, the computer-readable storage medium may be any tangible medium that contains or stores a program, and the program may be used by or in combination with an instruction execution system, apparatus, or device.

[0084] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meanings understood by those of ordinary skill in the field to which the present invention pertains. The "first", "second" and similar terms used in the present invention do not denote any order, quantity or importance, but are only used to distinguish different components. The terms such as "comprising" or "including" mean that the elements or items appearing before this word cover the elements or items listed after this word and their equivalents, without excluding other elements or items. The terms such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left", "right" are only used to indicate relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationships may also change accordingly.

[0085] As described above, these are only the preferred embodiments of the present invention, and are not intended to limit the present invention in any other form. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the technical solution content of the present invention still fall within the protection scope of the technical solution of the present invention.

[0086] This patent is not limited to the above best implementation manner. Anyone inspired by this patent can obtain various other forms of an interpretable assessment method for coastal erosion susceptibility based on random forests. All equal changes and modifications made according to the scope of the patent application of the present invention shall fall within the scope covered by this patent.

Claims

1. A random forest-based interpretable assessment method for coastal erosion susceptibility, characterized by: Select disaster-prone factors and obtain disaster-prone factor data; extract coastline data, divide coastal erosion areas based on coastline changes, and obtain the distribution of coastal erosion areas; pre-process the disaster-prone factor data and coastline data to construct a coastal erosion data set and divide it into training set and test set in proportion; construct a coastal erosion susceptibility assessment model based on the random forest algorithm; The models were comprehensively evaluated and compared to screen out the optimal model. Based on the optimal random forest model obtained through screening, the contribution of each factor in coastal erosion was quantified through feature importance analysis, the factors contributing to coastal erosion were identified, and the model was interpretably analyzed using partial dependence and SHAP to explain the impact of different disaster-prone factors on model predictions.

2. The method for interpretable assessment of coastal erosion susceptibility based on random forests according to claim 1, characterized in that: The selected disaster-predisposing factors include: elevation, slope, soil properties, topography, land use type, population density, GDP, annual rainfall, sea level, significant wave height, wave period, wave direction, suspended sediment concentration, tidal range, wind speed and coastline type.

3. The method for interpretable assessment of coastal erosion susceptibility based on random forests according to claim 1, characterized in that: The extraction of coastline data, the division of coastal erosion areas based on coastline changes, and the acquisition of coastal erosion area distribution are specifically as follows: based on remote sensing satellite images, the coastline data of a given area is extracted based on ENVI; according to the change in the position of the coastline in two adjacent periods, the area where the coastline in the latter period has retreated toward the shore relative to the previous period is identified as an erosion area, and the area is sampled as an erosion point with a given grid size as the basic unit; the area where the coastline in the latter period has silted up to the sea relative to the previous period or the coastline has not changed is identified as a non-erosion area, and the area is sampled as a non-erosion point; thereby the distribution of coastal erosion areas is obtained, and a historical inventory of coastal erosion in the study area is constructed.

4. The method for interpretable assessment of coastal erosion susceptibility based on random forests according to claim 1, characterized in that: The preprocessing is specifically as follows: the acquired disaster-prone factor data are processed through interpolation and resampling, and all factor resolutions are resampled into vector and raster data with a unified spatial resolution scale and unified projection coordinates consistent with the basic unit, and the Kriging method is used to interpolate and supplement the areas with missing data, and then the multicollinearity detection and data standardization of the disaster-prone factors are performed; the extracted coastline data are smoothed.

5. The method for interpretable assessment of coastal erosion susceptibility based on random forests according to claim 4, characterized in that: The smoothing of the extracted coastline is specifically as follows: establishing a model for monitoring the trend of coastline retreat rate using a digital coastline analysis system module on an ArcGIS platform; and using the end point change rate and the linear regression change rate as indicators when calculating the change rate of the coastline.

6. The method for interpretable assessment of coastal erosion susceptibility based on random forests according to claim 1, characterized in that: The construction of the coastal erosion susceptibility evaluation model based on the random forest algorithm is specifically: based on the random forest, a binary classification problem of coastal erosion: erosion / non-erosion is studied and analyzed.

7. The method for interpretable assessment of coastal erosion susceptibility based on random forests according to claim 1, characterized in that: The comprehensive evaluation and comparison of the models to screen out the optimal model specifically includes: based on the test data set, using accuracy, precision, recall rate, F1 score and Kappa index to evaluate the effectiveness of the model; based on the test data set, using the receiver operating characteristic curve ROC to evaluate the performance of the model.

8. The method for interpretable assessment of coastal erosion susceptibility based on random forests according to claim 1, characterized in that: After identifying the factors that contribute to coastal erosion, the importance of each factor is ranked; the model is interpreted through partial dependence and SHAP to explain and quantify the characteristic factors that induce coastal erosion.

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 an interpretable coastal erosion susceptibility assessment method based on random forests as described in any one of claims 1 to 8 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 the random forest-based interpretable coastal erosion susceptibility assessment method as described in any one of claims 1 to 8 are implemented.

Images