Gradient boosting regression-based mountain torrent ecological environment impact analysis method and system

Abstract

The present application relates to mountain torrent data processing technical field, especially to mountain torrent ecological environment influence analysis method and system based on gradient boosting regression, the present application introduces the nonlinear enhancement expression of ecological acceleration response factor, effectively solves the problem that the linear superposition hypothesis of traditional gradient boosting regression model cannot depict the acceleration deterioration trend of mountain torrent ecological loss, the hyperbolic tangent function is used for nonlinear modulation to the basic prediction result, the linear characteristic is kept when the prediction value is low, and the acceleration response is activated when the prediction value increases, so that the model can accurately simulate the explosive growth characteristics of ecological loss in strong mountain torrent or ecological fragile area. This improvement significantly improves the identification ability of the model to extreme ecological risk scenarios, avoids the problem that the prediction value of the traditional method is low in the high loss area, while maintaining the stability and computational efficiency of the gradient boosting framework, and provides a more natural evolution rule for the post-disaster ecological risk assessment of mountain torrents.

Description

Technical Field

[0001] This invention relates to the field of flash flood data processing technology, and in particular to a method and system for analyzing the ecological and environmental impacts of flash floods based on gradient boosting regression. Background Technology

[0002] Flash floods are special types of torrents that occur in mountainous areas, caused by torrential rain, snowmelt, or other water sources. They are short-lived, rapidly rising and falling, carrying large amounts of mud, sand, and rocks, and possess immense destructive power. The analysis of the formation mechanisms, risk assessment, and response strategies of flash flood disasters have been a research hotspot in the international field of disaster science. The formation and evolution of flash flood risk is the result of a complex interplay of climate, geomorphology, and human activities. Current research has shifted from analyzing single natural factors to multi-dimensional comprehensive assessments, with particular attention to the synergistic effects of urbanization and climate change. Global warming is exacerbating flood disasters by altering precipitation patterns and the frequency of extreme events, while rapid urbanization, leading to natural surface hardening, green space fragmentation, and the expansion of impermeable surfaces, significantly alters surface hydrological processes, further amplifying the risk of flash floods. Current research mainly focuses on the impacts of climate change and human activities on floods, assessing urban and social flood risks and vulnerabilities, and improving urban flood control capabilities. Research on the long-term impacts of floods on the ecological environment, risk warnings, and differentiated prevention measures remains limited.

[0003] Gradient boosting regression is an ensemble learning method commonly used in ecological and environmental impact assessment. However, flash floods have a significant nonlinear cumulative effect on the ecological environment, and the degree of ecological damage often accelerates with the increase of flash flood intensity. The traditional linear superposition gradient boosting regression model assumes that the contributions of each weak regressor to the final output are linearly added, which cannot characterize this nonlinear accelerated response feature. As a result, in strong flash floods or ecologically fragile areas, the model's prediction of ecological loss is too low, making it difficult to truly reflect the ecological risk under extreme scenarios.

[0004] Meanwhile, in traditional gradient boosting regression models, a uniform fixed learning rate η is used for all features and all samples when updating the model. However, the impact of flash floods on the ecological environment has significant nonlinearity and factor sensitivity differences. The fixed learning rate cannot distinguish the differences in the contributions of different ecological factors in model training, resulting in insufficient learning of highly sensitive ecological factors and overfitting to low-sensitivity factors, thereby reducing the accuracy of ecological impact assessment. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method and system for analyzing the ecological and environmental impacts of flash floods based on gradient boosting regression, which solves the problems existing in the prior art.

[0006] This invention provides a method for analyzing the ecological and environmental impacts of flash floods based on gradient boosting regression. The method includes the following steps: S1: Acquire data from the study area and perform data preprocessing operations; S2: Construct a gradient boosting regression model for analyzing the ecological and environmental impact of flash floods; The expression for the gradient boosting regression model is: ; In the formula, For the input feature vector, The output value of the gradient boosting regression model is... For the m-th regressor, The weight coefficients of the m-th regressor are... For the ecological acceleration intensity coefficient, Here, ∑ represents the response sensitivity coefficient, tanh represents the summation term, and tanh represents the hyperbolic tangent activation function. S3; The gradient boosting regression model is trained using an adaptive ecological sensitivity factor weighted learning rate. Specifically, S3 consists of: S3.1: Calculating the ecological sensitivity weights for each feature dimension; S3.2: Constructing an adaptive learning rate based on the ecological sensitivity weights for each feature dimension; S3.3: Training the gradient boosting regression model based on the adaptive learning rate. S4; Conduct an assessment of the ecological and environmental impact of flash floods based on the gradient boosting regression model.

[0007] Preferably, the ecological acceleration intensity coefficient and the response sensitivity coefficient are hyperparameters obtained by training the gradient boosting regression model.

[0008] Preferably, the loss function L of the gradient boosting regression model is: ; In the formula, y is the target variable for ecological impact.

[0009] Preferably, in S3.1, the ecological sensitivity weight of each dimension feature... The expression is: ; In the formula, Let be the Pearson correlation coefficient between the j-th dimension feature and the ecological impact target variable. The function for calculating the Pearson correlation coefficient. denoted as the Pearson correlation coefficient between the k-th feature and the target variable of ecological impact, where d is the feature dimension, with a value of 5, representing five dimensions: rainfall factor data, hydrological factor data, topographic factor data, vegetation factor data, and soil factor data.

[0010] Preferably, in step S3.2, the expression for the adaptive learning rate is: ; In the formula, Let be the adaptive learning rate for the j-th dimension feature. Based on the learning rate.

[0011] Preferably, step S3.3 specifically involves: initializing the gradient boosting regression model before starting iterative training; The training set is input into the gradient boosting regression model for training. The residual of the m-th round is calculated. After completing the previous round of model training, the m-th round first needs to calculate the difference between the actual ecological impact value of each sample and the current model prediction value, i.e., the residual. The expression is: ; In the formula, Let m be the residual of the m-th iteration. Let be the target variable for the ecological impact of the i-th sample, i.e., the actual value. These are the model predictions from the previous round; Construct a feature-level adaptive learning rate matrix; organize the learning rates of all features into a diagonal matrix form, expressed as: ; In the formula, Let be the learning rate matrix for the m-th iteration. Let be the adaptive learning rate for the d-th dimension feature; To fit a residual regression tree, the m-th regression tree is trained using the residuals weighted by the adaptive learning rate as the fitting target. ; Optimal step size estimation: After obtaining the output of the m-th regression tree, it is necessary to determine the contribution weight of the regression tree in the final model, i.e., the step size of the m-th regression tree. The expression is: ; In the formula, The step size of the regression tree, These are the model predictions from the previous round. Let be the predicted value of the m-th regression tree for the i-th sample; Model updates, after step size estimation, incorporate the newly trained regression tree into the overall model to progressively enhance its capabilities. The expression is as follows: ; Repeat the above steps until any of the following termination conditions are met: the preset maximum number of iterations M is reached; the validation set loss no longer decreases or increases for T consecutive rounds; the improvement in model performance from adding a regression tree is less than a set threshold. Once the termination conditions are met, the training process ends, and the final mountain torrent ecological environment impact assessment model is obtained.

[0012] Preferably, in step S1, the spatial range Ω of the study area is determined based on historical flash flood disaster records, watershed boundaries, and administrative divisions.

[0013] Preferably, rainfall factor data, hydrological factor data, topographic factor data, vegetation factor data, and soil factor data are collected for the study area; wherein, the rainfall factor data is the maximum hourly rainfall of a single event obtained from meteorological stations, the hydrological factor data is the peak flow rate obtained from hydrological stations, the topographic factor data is the average slope of the terrain extracted from a DEM model, the vegetation factor data is the vegetation cover obtained from remote sensing NDVI inversion, and the soil factor data is the map erosion modulus obtained from soil survey data; Simultaneously, ecological environment status data of the study area before and after the flash flood were obtained, and the target variable y of ecological impact was calculated, with the expression as follows: ; In the formula, Indicates the value of ecosystem services prior to flash floods. This indicates the value of ecosystem services after a flash flood.

[0014] Preferably, the data preprocessing operations include data cleaning and outlier handling, data normalization, and feature construction.

[0015] According to another aspect of the present invention, a method system for analyzing the ecological and environmental impacts of flash floods based on gradient boosting regression is provided. The system employs the aforementioned method for analyzing the ecological and environmental impacts of flash floods based on gradient boosting regression, and the system includes: The data acquisition module is used to acquire data from the study area and perform data preprocessing operations. The model building module is used to build gradient boosting regression models for analyzing the ecological and environmental impacts of flash floods. The model training module is used to train the gradient boosting regression model using an adaptive ecological sensitivity factor weighted learning rate. The assessment module is used to assess the ecological and environmental impact of flash floods based on the gradient boosting regression model.

[0016] The present invention has the following technical effects: This embodiment effectively addresses the problem that traditional gradient boosting regression models, by introducing a nonlinear enhancement expression for the ecological acceleration response factor, cannot capture the accelerated deterioration trend of flash flood ecological losses due to the linear superposition assumption. It utilizes a hyperbolic tangent function to nonlinearly modulate the basic prediction results, maintaining linearity when prediction values ​​are low and activating the accelerated response as prediction values ​​increase. This allows the model to accurately simulate the explosive growth characteristics of ecological losses in strong flash floods or ecologically fragile areas. This improvement significantly enhances the model's ability to identify extreme ecological risk scenarios, avoids the problem of traditional methods underestimating high-loss areas, and maintains the stability and computational efficiency of the gradient boosting framework. It provides a more natural and evolutionary approach for post-flash flood ecological risk assessment.

[0017] This embodiment effectively addresses the ecological impact assessment bias caused by the uniform fixed learning rate in traditional gradient boosting regression models by introducing an adaptive ecological sensitivity factor weighted learning rate. Sensitive weights are constructed by calculating the Pearson correlation coefficients between each ecological factor and the target variable, and the learning rate for different features is dynamically adjusted. This allows ecological factors sensitive to flash flood responses, such as vegetation cover and soil erosion, to receive larger update steps, thereby accelerating the model's fitting speed to these key influencing factors and improving the accuracy of identifying ecologically vulnerable areas. Simultaneously, the learning rate is automatically reduced for non-sensitive factors, effectively suppressing noise interference and overfitting, enhancing the model's generalization ability. Without altering the overall framework of the gradient boosting algorithm, differentiated learning at the feature level is achieved, enabling the model to more accurately quantify the nonlinear damage of flash floods to the ecosystem. This improves the scientific rigor and reliability of the assessment results, providing more precise data support for post-flash flood ecological restoration and disaster prevention and mitigation decisions. Attached Figure Description

[0018] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a flowchart of the method for analyzing the ecological and environmental impact of flash floods based on gradient boosting regression provided in this embodiment of the invention; Figure 2 This is a flowchart of constructing an adaptive learning rate based on the ecological sensitivity weights of each feature dimension, provided in an embodiment of the present invention. Detailed Implementation

[0020] 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. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0021] Example 1, Figure 1 The flowchart of the method for analyzing the ecological and environmental impact of flash floods based on gradient boosting regression is shown, as follows: Figure 1 As shown, the method for analyzing the ecological and environmental impacts of flash floods based on gradient boosting regression includes the following steps: S1: Acquire data from the study area and perform data preprocessing operations; Based on historical records of flash floods, watershed boundaries, and administrative divisions, the spatial scope Ω of the study area was determined.

[0022] Rainfall factor data, hydrological factor data, topographic factor data, vegetation factor data, and soil factor data were collected for the study area. Among them, the rainfall factor data was the maximum hourly rainfall observed by meteorological stations, the hydrological factor data was the peak flow observed by hydrological stations, the topographic factor data was the average slope of the terrain extracted by the DEM model, the vegetation factor data was the vegetation coverage obtained by remote sensing NDVI inversion, and the soil factor data was the map erosion modulus obtained by soil survey data.

[0023] Simultaneously, ecological environment status data of the study area before and after the flash flood were obtained, and the target variable y of ecological impact was calculated, with the expression as follows: ; In the formula, Indicates the value of ecosystem services prior to flash floods. This indicates the value of ecosystem services after a flash flood.

[0024] The data preprocessing operations include data cleaning and outlier handling, data normalization, and feature construction. Data cleaning and outlier handling include: removing evaluation units with a missing value rate exceeding 30%, and identifying and correcting outliers using box plots; the data normalization process involves normalizing the data from the study area using a maximum-minimum normalization method.

[0025] The feature construction involves transforming the normalized data into a feature vector.

[0026] S2: Construct a gradient boosting regression model for analyzing the ecological and environmental impact of flash floods; Gradient boosting regression is an ensemble learning method that, by stacking multiple weak regressors, gradually reduces the residual between the model's predicted values ​​and the actual ecological impact values, thereby achieving a fit to complex nonlinear relationships. In the traditional gradient boosting regression model, the expression is: ; In the formula, For the input feature vector, The output value of the gradient boosting regression model represents the degree of impact of flash floods on the ecological environment. For the k-th weak regressor, is the weight coefficient of the k-th regressor, and K is the total number of weak regressors.

[0027] However, flash floods have a significant nonlinear cumulative effect on the ecological environment. The degree of ecological damage often accelerates as the intensity of flash floods increases. Traditional linear superposition models assume that the contributions of each weak regressor to the final output are linearly cumulative, which cannot characterize this nonlinear accelerated response feature. As a result, in strong flash floods or ecologically fragile areas, the model's prediction of ecological loss is too low and it is difficult to truly reflect the ecological risk under extreme scenarios.

[0028] To address the aforementioned issues, this step improves the expression of the traditional gradient boosting regression model by introducing an ecological acceleration response factor on top of the linear superposition structure, enabling the model to express the accelerating deterioration trend of ecological loss.

[0029] Specifically, the expression for the gradient boosting regression model is: ; In the formula, For the input feature vector, The output value of the gradient boosting regression model represents the degree of impact of flash floods on the ecological environment. For the m-th regressor, The weight coefficients of the m-th regressor are... This is the ecological acceleration intensity coefficient, used to control the magnitude of nonlinear enhancement. ∑ is the response sensitivity coefficient used to adjust the trigger threshold of the acceleration effect, ∑ is the summation term, and tanh is the hyperbolic tangent activation function.

[0030] It is worth emphasizing that the ecological acceleration intensity coefficient and the response sensitivity coefficient are hyperparameters, obtained by training the gradient boosting regression model using training samples.

[0031] This embodiment effectively addresses the problem that traditional gradient boosting regression models, by introducing a nonlinear enhancement expression for the ecological acceleration response factor, cannot capture the accelerated deterioration trend of flash flood ecological losses due to the linear superposition assumption. It utilizes a hyperbolic tangent function to nonlinearly modulate the basic prediction results, maintaining linearity when prediction values ​​are low and activating the accelerated response as prediction values ​​increase. This allows the model to accurately simulate the explosive growth characteristics of ecological losses in strong flash floods or ecologically fragile areas. This improvement significantly enhances the model's ability to identify extreme ecological risk scenarios, avoids the problem of traditional methods underestimating high-loss areas, and maintains the stability and computational efficiency of the gradient boosting framework. It provides a more natural and evolutionary approach for post-flash flood ecological risk assessment.

[0032] Furthermore, to quantitatively describe the model's prediction error, the loss function is defined as the mean squared error, i.e., the loss function of the gradient boosting regression model is: ; In the formula, y is the target variable for ecological impact.

[0033] S3; The gradient boosting regression model is trained using an adaptive ecological sensitivity factor weighted learning rate. In traditional gradient boosting regression models, a uniform fixed learning rate η is used for all features and all samples during model updates. However, the impact of flash floods on the ecological environment has significant nonlinearity and factor sensitivity differences. The fixed learning rate cannot distinguish the differences in the contributions of different ecological factors in model training, resulting in insufficient learning of highly sensitive ecological factors and overfitting to low-sensitivity factors, thereby reducing the accuracy of ecological impact assessment.

[0034] This embodiment proposes an adaptive ecological sensitivity factor weighted learning rate. In each iteration, the learning rate is dynamically adjusted based on the correlation strength between different ecological factors and the target variable, thereby achieving differentiated updates. Specifically, as shown... Figure 2 As shown, S3 specifically includes: S3.1: Calculate the ecological sensitivity weight for each feature dimension; Before training the model, the ecological sensitivity weights for each feature dimension are first calculated. The expression is: ; In the formula, Let be the Pearson correlation coefficient between the j-th dimension feature and the ecological impact target variable. The function for calculating the Pearson correlation coefficient. Let d be the Pearson correlation coefficient between the k-th feature and the target variable of ecological impact, and d be the feature dimension. In this embodiment, d is set to 5, which represents the five dimensions: rainfall factor data, hydrological factor data, topographic factor data, vegetation factor data, and soil factor data.

[0035] S3.2: Construct an adaptive learning rate based on the ecological sensitivity weights of each feature dimension; In this step, the traditional fixed learning rate η is expanded into a function of the feature dimension, expressed as: ; In the formula, Let be the adaptive learning rate for the j-th dimension feature. Based on the learning rate.

[0036] S3.3: Train the gradient boosting regression model according to the adaptive learning rate; Before starting iterative training, the gradient boosting regression model needs to be initialized. Since this embodiment uses mean squared error as the loss function, theoretically the optimal initial model should minimize the sum of the prediction errors of all samples. It can be deduced that under mean squared error loss, the optimal initial value is the arithmetic mean of the ecological impact target values ​​of all samples in the training set.

[0037] The training set is input into the gradient boosting regression model for training. The residual of the m-th round is calculated. After completing the previous round of model training, the m-th round first needs to calculate the difference between the actual ecological impact value of each sample and the current model prediction value, i.e., the residual. In this step, the residual is defined as the negative gradient of the loss function with respect to the predicted value, expressed as: ; In the formula, The residual is the value of the m-th iteration, which reflects the amount of information that the current model has not yet explained. The larger the absolute value of the residual, the greater the prediction bias of the current model for this sample. The sign of the residual indicates whether the predicted value is too high or too low. This residual will serve as the fitting target for the next regression tree, guiding the model to continuously approach the true ecological impact value. Let be the target variable for the ecological impact of the i-th sample, i.e., the actual value. These are the model predictions from the previous round.

[0038] Construct a feature-level adaptive learning rate matrix; organize the learning rates of all features into a diagonal matrix form, expressed as: ; In the formula, Let m be the learning rate matrix for the m-th iteration. is the adaptive learning rate for the d-th dimension feature.

[0039] The feature-level adaptive learning rate matrix is ​​used to apply different update intensities to different features. High-sensitivity ecological factors obtain a larger effective learning rate, which accelerates the model's capture of their patterns, while low-sensitivity factors are suppressed, reducing noise interference and the risk of overfitting.

[0040] To fit a residual regression tree, the m-th regression tree is trained using the residuals weighted by the adaptive learning rate as the fitting target. Regression trees recursively divide the feature space into several non-overlapping regions and output constant prediction values ​​in each region, thereby modeling non-linear relationships.

[0041] Optimal step size estimation: After obtaining the output of the m-th regression tree, it is necessary to determine the contribution weight of the regression tree in the final model, i.e., the step size of the m-th regression tree. The step size should be chosen to minimize the overall loss of the model after adding the new tree, and the expression is: ; In the formula, The step size of the regression tree, These are the model predictions from the previous round. Let be the predicted value of the m-th regression tree for the i-th sample.

[0042] The optimal step size obtained through linear search can prevent a single tree from causing excessively large or small perturbations to the model, thereby ensuring the stability of the training process.

[0043] Model updates, after step size estimation, incorporate the newly trained regression tree into the overall model to progressively enhance its capabilities. The expression is as follows: .

[0044] Compared with traditional update methods, this step implicitly incorporates the role of feature-level adaptive learning rate during the construction stage of the regression tree. Therefore, this update process can more accurately characterize the impact intensity of flash floods on different ecological factors, and is especially suitable for mountainous watersheds with strong ecological heterogeneity and significant differences in factor responses.

[0045] Repeat the above steps until any of the following termination conditions are met: the preset maximum number of iterations M is reached; the validation set loss no longer decreases or increases for T consecutive rounds; the improvement in model performance from adding a regression tree is less than a set threshold.

[0046] Once the termination conditions are met, the training process ends, and the final mountain torrent ecological environment impact assessment model is obtained.

[0047] This embodiment effectively addresses the ecological impact assessment bias caused by the uniform fixed learning rate in traditional gradient boosting regression models by introducing an adaptive ecological sensitivity factor weighted learning rate. Sensitive weights are constructed by calculating the Pearson correlation coefficients between each ecological factor and the target variable, and the learning rate for different features is dynamically adjusted. This allows ecological factors sensitive to flash flood responses, such as vegetation cover and soil erosion, to receive larger update steps, thereby accelerating the model's fitting speed to these key influencing factors and improving the accuracy of identifying ecologically vulnerable areas. Simultaneously, the learning rate is automatically reduced for non-sensitive factors, effectively suppressing noise interference and overfitting, enhancing the model's generalization ability. Without altering the overall framework of the gradient boosting algorithm, differentiated learning at the feature level is achieved, enabling the model to more accurately quantify the nonlinear damage of flash floods to the ecosystem. This improves the scientific rigor and reliability of the assessment results, providing more precise data support for post-flash flood ecological restoration and disaster prevention and mitigation decisions.

[0048] S4; Conduct an ecological and environmental impact assessment of flash floods based on the aforementioned gradient boosting regression model; The real-time monitoring data of the area to be evaluated is input into the trained gradient boosting regression model, which outputs the ecological environment impact index.

[0049] Example 2: This invention also provides a method and system for analyzing the ecological and environmental impacts of flash floods based on gradient boosting regression. The system employs the method for analyzing the ecological and environmental impacts of flash floods based on gradient boosting regression from Example 1. The system includes: The data acquisition module is used to acquire data from the study area and perform data preprocessing operations. The model building module is used to build gradient boosting regression models for analyzing the ecological and environmental impacts of flash floods. The model training module is used to train the gradient boosting regression model using an adaptive ecological sensitivity factor weighted learning rate. The assessment module is used to assess the ecological and environmental impact of flash floods based on the gradient boosting regression model.

[0050] Example 3: The present invention also provides an electronic device, including one or more processors and a memory.

[0051] A processor can be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and can control other components in an electronic device to perform desired functions.

[0052] The memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and a processor may execute the program instructions to implement the gradient-enhancing regression-based flash flood ecological environment impact analysis method described above in any embodiment of this application, and / or other desired functions. Various contents such as initial extrinsic parameters and thresholds may also be stored in the computer-readable storage medium.

[0053] 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 or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for analyzing the ecological and environmental impacts of flash floods based on gradient boosting regression, characterized in that, The method includes the following steps: S1: Acquire data from the study area and perform data preprocessing operations; S2: Construct a gradient boosting regression model for analyzing the ecological and environmental impact of flash floods; The expression for the gradient boosting regression model is: ； In the formula, For the input feature vector, To boost the output of the gradient-boosting regression model, Let m be the predicted value of the m-th regression tree. Let m be the step size of the m-th regression tree. For the ecological acceleration intensity coefficient, Here, ∑ represents the response sensitivity coefficient, tanh represents the summation term, and tanh represents the hyperbolic tangent activation function. S3: The gradient boosting regression model is trained using an adaptive ecological sensitivity factor weighted learning rate; S3.1: Calculate the ecological sensitivity weight for each feature dimension; the ecological sensitivity weight for each feature dimension. The expression is: ； In the formula, Let be the Pearson correlation coefficient between the j-th dimension feature and the ecological impact target variable. The function for calculating the Pearson correlation coefficient. Let be the Pearson correlation coefficient between the k-th feature and the ecological impact target variable, where d is the feature dimension, with a value of 5, representing five dimensions: rainfall factor data, hydrological factor data, topographic factor data, vegetation factor data, and soil factor data; S3.2: Construct an adaptive learning rate based on the ecological sensitivity weights of each feature dimension; the expression for the adaptive learning rate is: ； In the formula, Let be the adaptive learning rate for the j-th dimension feature. Base learning rate; S3.3: Train the gradient boosting regression model using an adaptive learning rate; specifically, initialize the gradient boosting regression model before starting iterative training. The training set is input into the gradient boosting regression model for training. The residuals of the m-th round are calculated. After completing the previous round of model training, the m-th round first needs to calculate the difference between the actual ecological impact value of each sample and the current model prediction value, i.e., the residual. The expression is: ； In the formula, Let m be the residual of the m-th iteration. Let be the target variable for the ecological impact of the i-th sample, i.e., the actual value. These are the model predictions from the previous round; Construct a feature-level adaptive learning rate matrix; organize the learning rates of all features into a diagonal matrix form, expressed as: ； In the formula, Let m be the learning rate matrix for the m-th iteration. is the adaptive learning rate for the d-th dimension feature; To fit a residual regression tree, the m-th regression tree is trained using the residuals weighted by the adaptive learning rate as the fitting target. ; Optimal step size estimation: After obtaining the output of the m-th regression tree, it is necessary to determine the contribution weight of the regression tree in the final model, i.e., the step size of the m-th regression tree. The expression is: ； In the formula, The step size of the regression tree, These are the model predictions from the previous round. Let be the predicted value of the m-th regression tree for the i-th sample; Model updates, after step size estimation, incorporate the newly trained regression tree into the overall model to progressively enhance its capabilities. The expression is as follows: ； Repeat the above steps until any of the following termination conditions are met: the preset maximum number of iterations M is reached; the validation set loss no longer decreases or increases for T consecutive rounds; the improvement in model performance from adding a regression tree is less than a set threshold. Once the termination conditions are met, the training process ends, and the final mountain torrent ecological environment impact assessment model is obtained. S4: Conduct an assessment of the ecological and environmental impact of flash floods based on the gradient boosting regression model.

2. The method for analyzing the ecological and environmental impacts of flash floods based on gradient boosting regression according to claim 1, characterized in that, The ecological acceleration intensity coefficient and the response sensitivity coefficient are hyperparameters, obtained by training the gradient boosting regression model.

3. The method for analyzing the ecological and environmental impacts of flash floods based on gradient boosting regression according to claim 1, characterized in that, The loss function L of the gradient boosting regression model is: ； In the formula, y is the target variable for ecological impact.

4. The method for analyzing the ecological and environmental impacts of flash floods based on gradient boosting regression according to claim 1, characterized in that, In S1, the spatial range Ω of the study area is determined based on historical flash flood disaster records, watershed boundaries, and administrative divisions.

5. The method for analyzing the ecological and environmental impact of flash floods based on gradient boosting regression according to claim 4, characterized in that, Rainfall factor data, hydrological factor data, topographic factor data, vegetation factor data, and soil factor data were collected for the study area. Among them, the rainfall factor data was the maximum hourly rainfall observed by meteorological stations, the hydrological factor data was the peak flow observed by hydrological stations, the topographic factor data was the average slope of the terrain extracted by the DEM model, the vegetation factor data was the vegetation coverage obtained by remote sensing NDVI inversion, and the soil factor data was the soil erosion modulus obtained by soil survey data. Simultaneously, ecological environment status data of the study area before and after the flash flood were obtained, and the target variable y of ecological impact was calculated, with the expression as follows: ； In the formula, Indicates the value of ecosystem services prior to flash floods. This indicates the value of ecosystem services after a flash flood.

6. The method for analyzing the ecological and environmental impact of flash floods based on gradient boosting regression according to claim 1, characterized in that, The data preprocessing operations include data cleaning and outlier handling, data normalization, and feature construction.

7. A system for analyzing the ecological and environmental impacts of flash floods based on gradient boosting regression, characterized in that, The system employs the gradient-enhancing regression-based method for analyzing the ecological and environmental impacts of flash floods as described in any one of claims 1-6, and the system comprises: The data acquisition module is used to acquire data from the study area and perform data preprocessing operations. The model building module is used to build gradient boosting regression models for analyzing the ecological and environmental impacts of flash floods. The model training module is used to train the gradient boosting regression model using an adaptive ecological sensitivity factor weighted learning rate. The assessment module is used to assess the ecological and environmental impact of flash floods based on the gradient boosting regression model.

Images