Typhoon disaster chain risk assessment method, system, equipment and medium
By conducting feature engineering and decision-making tree weighted prediction of typhoon-related data, combined with incremental learning methods, the dynamic adaptability problem of typhoon disaster chain risk assessment is solved, and efficient and accurate risk assessment and scientific disaster prevention strategy formulation are achieved.
Patent Information
- Application Number
- CN202510616466.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-14
- Publication Date
- 2025-08-19
AI Technical Summary
The existing typhoon disaster chain risk assessment methods cannot adapt to changes in dynamic data, resulting in inaccurate risk assessment and poor timeliness, making it difficult to meet the real-time assessment needs of the rapid evolution of disasters.
Feature engineering is carried out by obtaining typhoon-related data at the current moment, using decision trees and evaluation models for weighted predictions, determining the risk assessment level, and adjusting model parameters through incremental learning methods to adapt to dynamic data changes.
It improves the accuracy and timeliness of risk assessment, can quickly respond to dynamic data changes, capture the complex interactions and nonlinear relationships between various factors in the typhoon disaster chain, and ensures the scientificity and rationality of disaster prevention strategies.
Smart Images
Figure CN120508981A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of risk assessment, and in particular to a typhoon disaster chain risk assessment method, system, equipment and medium. Background Art
[0002] The typhoon disaster chain refers to direct disasters such as strong winds, heavy rains, storm surges caused by typhoons, as well as secondary disasters such as landslides and floods caused by them. The occurrence of this disaster chain is dynamic, complex and regional. At the same time, factors such as the typhoon's movement path, intensity changes and interaction with the terrain make the evolution process of the disaster chain full of uncertainty and difficult to predict accurately.
[0003] Traditional typhoon disaster chain risk assessment methods primarily include complex network analysis and Bayesian network methods, but both have their limitations. While complex network methods can characterize the topological relationships between hazard elements, they use static graph structures to represent hazard propagation paths and fail to capture the nonlinear characteristics of hazard chains over time. Therefore, when hazard chains involve dynamic changes or data integrity is insufficient, these methods often miss key information, leading to inaccurate risk assessments. While Bayesian networks can handle uncertainty, their inference accuracy is highly dependent on prior probability distributions. The complexity and regional variability of typhoon disaster chains make constructing universal probability tables difficult. Furthermore, in disaster emergency response scenarios, Bayesian network parameter updates require a complete data collection-model training-validation cycle, making them difficult to adapt to the real-time assessment requirements of rapidly evolving disasters. Therefore, developing a typhoon disaster chain risk assessment method that can adapt to dynamic data changes is particularly urgent. This will help improve the accuracy and timeliness of risk assessments and provide stronger support for disaster prevention and mitigation efforts in coastal areas. Summary of the Invention
[0004] The present invention provides a typhoon disaster chain risk assessment method, system, equipment and medium, which can adapt to the typhoon disaster chain risk assessment with dynamic data changes and improve the accuracy and timeliness of risk assessment.
[0005] The present invention provides a typhoon disaster chain risk assessment method, comprising:
[0006] Acquire first typhoon-related data at the current moment, and perform feature engineering on the first typhoon-related data to obtain multidimensional features;
[0007] The multidimensional features are input into a preset evaluation model, and the key risk factors in the multidimensional features are determined based on the feature importance weights. The key risk factors are weightedly predicted through each decision tree to obtain a comprehensive prediction value of multiple decision trees, and the risk assessment level is determined based on the comprehensive prediction value, wherein the feature importance weights are determined during the training process of the evaluation model, and each of the decision trees is obtained by residual fitting in the process of training the evaluation model using historical typhoon data. The evaluation model is determined based on the model parameters optimized according to the prediction values of each decision tree.
[0008] The embodiment of the present application ensures the timeliness of the data by real-time acquisition and processing of typhoon data at the current moment, which facilitates subsequent risk assessment based on the latest data; by performing feature engineering on the first typhoon-related data, the first typhoon-related data can be converted into multi-dimensional features that can effectively characterize the risk of the typhoon disaster chain, and the most valuable information for risk assessment can be quickly extracted from the first typhoon-related data, which facilitates subsequent risk assessment; by inputting the multi-dimensional features into a preset assessment model, the processing and prediction of a large amount of data can be completed in a relatively short period of time, meeting the demand for rapid response under dynamic data changes, and at the same time effectively capturing the complex interactions between various factors in the typhoon disaster chain. The mutual and nonlinear relationship significantly improves the prediction accuracy of risk assessment; the key risk factors in the multidimensional features are determined by using the feature importance weights determined by the evaluation model during the training process, which can reduce unnecessary feature interference and improve the sensitivity of the model to key factors, thereby improving the accuracy of risk assessment; by making weighted predictions on the key risk factors through each decision tree, the prediction can be completed in a shorter time, ensuring the efficiency of the entire evaluation process. At the same time, the weighted prediction integrates the results of multiple decision trees, which can better fit the actual distribution of the data and improve the accuracy of the prediction; the risk assessment level is determined based on the comprehensive prediction value, which can more accurately reflect the risk level of the current typhoon disaster chain. Compared with the existing technology, the present application can adapt to the risk assessment of typhoon disaster chains with dynamic data changes, and improve the accuracy and timeliness of risk assessment.
[0009] Furthermore, the feature engineering is performed on the first typhoon-related data to obtain multi-dimensional features, specifically:
[0010] Performing data cleaning on the typhoon-related data to obtain a first processing result;
[0011] Normalizing the first processing result to obtain a second processing result;
[0012] Feature extraction and feature construction are performed on the second processing result to obtain multi-dimensional features.
[0013] In this way, by performing feature engineering on the first typhoon-related data, the first typhoon-related data can be converted into multi-dimensional features that can effectively characterize the risk of the typhoon disaster chain. The most valuable information for risk assessment can be quickly extracted from the first typhoon-related data, facilitating subsequent risk assessment.
[0014] Furthermore, the training process of the evaluation model is specifically as follows:
[0015] Acquire the historical typhoon data of a preset time, and perform feature engineering on the historical typhoon data to obtain a training data set;
[0016] An initial evaluation model is constructed, and the training data set is input into the initial evaluation model. In each iteration, a pseudo residual is calculated based on the current prediction value of the initial evaluation model, and a decision tree is fitted using the pseudo residual as the target variable. The prediction value corresponding to the current decision tree is added to the initial evaluation model to optimize the first model parameters until a preset iteration termination condition is met, and an evaluation model is determined based on the target model parameters.
[0017] In this way, by training the evaluation model, it is convenient for subsequent evaluation models to complete the processing and prediction of large amounts of data in a shorter period of time, meeting the needs of rapid response under dynamic data changes, while effectively capturing the complex interactions and nonlinear relationships between various factors in the typhoon disaster chain, and significantly improving the prediction accuracy of risk assessment.
[0018] Furthermore, the calculation formula for adding the predicted value of the current decision tree to the initial evaluation model to optimize the first model parameter is specifically:
[0019] F m (x) = F m-1 (x)+γ m h m (x);
[0020]
[0021] Where, F m (x) and F m-1 (x) are the overall predicted values for the mth and m-1th iterations, γ is the overall model parameter, and γ m is the model parameter determined in the mth time, h m (x) is the predicted value corresponding to the decision tree fitted at the mth iteration, y i is the true value of the i-th sample, x i The training data of the i-th sample, n is the total number of training samples.
[0022] Furthermore, the key risk factors are weighted predicted by each decision tree to obtain a comprehensive prediction value of multiple decision trees, specifically:
[0023] For each decision tree, make judgments from the root node to the leaf node based on the key risk factors and output several initial prediction values;
[0024] The initial prediction values are weighted and summed according to the model weights to obtain the comprehensive prediction value of multiple decision trees.
[0025] In this way, by performing weighted prediction on the key risk factors through each decision tree, the prediction can be completed in a shorter time, ensuring the efficiency of the entire evaluation process. At the same time, the weighted prediction integrates the results of multiple decision trees, which can better fit the actual distribution of the data and improve the accuracy of the prediction.
[0026] Furthermore, after determining the risk assessment level based on the comprehensive prediction value, the method further includes:
[0027] Get the second typhoon related data at the current moment;
[0028] Inputting the second typhoon-related data into the evaluation model, and adjusting the target model parameters of the evaluation model through an incremental learning method to obtain second model parameters;
[0029] Based on the second model parameters, an updated risk assessment level is determined.
[0030] In this way, by adjusting the target model parameters of the assessment model through the incremental learning method, it can be ensured that the model can adjust the disaster risk assessment level according to the latest dynamics of the typhoon.
[0031] Furthermore, after determining the risk assessment level based on the comprehensive prediction value, the method further includes:
[0032] Determining a corresponding first disaster prevention strategy based on the risk assessment level according to a preset first matching rule;
[0033] Determining a corresponding second disaster prevention strategy based on the key risk factors according to a preset second matching rule;
[0034] A target disaster prevention strategy is determined based on the first disaster prevention strategy and the second disaster prevention strategy, and disaster prevention is performed according to the target disaster prevention strategy.
[0035] In this way, by determining the target disaster prevention strategy through risk assessment levels and key risk factors, we can ensure that disaster prevention measures are both comprehensive and targeted, and improve the scientific nature and rationality of disaster prevention strategies.
[0036] Another embodiment of the present invention further provides a typhoon disaster chain risk assessment system, comprising: an acquisition module and an assessment module;
[0037] The acquisition module is used to acquire the first typhoon-related data at the current moment, and perform feature engineering on the first typhoon-related data to obtain multidimensional features;
[0038] The evaluation module is used to input the multidimensional features into a preset evaluation model, determine the key risk factors in the multidimensional features based on the feature importance weights, and perform weighted predictions on the key risk factors through each decision tree to obtain a comprehensive prediction value of multiple decision trees, and determine the risk assessment level based on the comprehensive prediction value, wherein the feature importance weights are determined during the training process of the evaluation model, and each of the decision trees is obtained by residual fitting in the process of training the evaluation model using historical typhoon data. The evaluation model is determined based on the model parameters optimized according to the prediction values of each decision tree.
[0039] The embodiment of the present application ensures the timeliness of the data by real-time acquisition and processing of typhoon data at the current moment, which facilitates subsequent risk assessment based on the latest data; by performing feature engineering on the first typhoon-related data, the first typhoon-related data can be converted into multi-dimensional features that can effectively characterize the risk of the typhoon disaster chain, and the most valuable information for risk assessment can be quickly extracted from the first typhoon-related data, which facilitates subsequent risk assessment; by inputting the multi-dimensional features into a preset assessment model, the processing and prediction of a large amount of data can be completed in a relatively short period of time, meeting the demand for rapid response under dynamic data changes, and at the same time effectively capturing the complex interactions between various factors in the typhoon disaster chain. The mutual and nonlinear relationship significantly improves the prediction accuracy of risk assessment; the key risk factors in the multidimensional features are determined by using the feature importance weights determined by the evaluation model during the training process, which can reduce unnecessary feature interference and improve the sensitivity of the model to key factors, thereby improving the accuracy of risk assessment; by making weighted predictions on the key risk factors through each decision tree, the prediction can be completed in a shorter time, ensuring the efficiency of the entire evaluation process. At the same time, the weighted prediction integrates the results of multiple decision trees, which can better fit the actual distribution of the data and improve the accuracy of the prediction; the risk assessment level is determined based on the comprehensive prediction value, which can more accurately reflect the risk level of the current typhoon disaster chain. Compared with the existing technology, the present application can adapt to the risk assessment of typhoon disaster chains with dynamic data changes, and improve the accuracy and timeliness of risk assessment.
[0040] Another embodiment of the present invention also provides a terminal device, including: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the steps of the typhoon disaster chain risk assessment method of the present invention are implemented.
[0041] Another embodiment of the present invention further provides a computer-readable storage medium, comprising: a stored computer program, which controls the device where the computer-readable storage medium is located to execute the steps of the typhoon disaster chain risk assessment method of the present invention when the computer program is running. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for use in the implementation. Obviously, the drawings described below are only some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0043] Figure 1 This is a flow chart of an embodiment of the typhoon disaster chain risk assessment method provided by the present application;
[0044] Figure 2 This is a structural diagram of an embodiment of the typhoon disaster chain risk assessment system provided by this application. DETAILED DESCRIPTION
[0045] To make the objectives, technical solutions, and advantages of this application more clear, the technical solutions in this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.
[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.
[0047] In the description of the embodiments of this application, the technical terms "first" and "second" are used only to distinguish different objects and should not be understood to indicate or imply relative importance or implicitly specify the quantity, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, the meaning of "plurality" is more than two, unless otherwise clearly and specifically defined.
[0048] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0049] In the description of the embodiments of this application, the term "and / or" is simply a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent the following three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.
[0050] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).
[0051] In the description of the embodiments of the present application, unless otherwise expressly specified or limited, technical terms such as "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integration; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; internal connections between two components or interactions between two components. Those skilled in the art can understand the specific meanings of the above terms in the embodiments of the present application based on specific circumstances.
[0052] See also Figure 1 In order to solve the problem that the existing technology cannot adapt to the typhoon disaster chain risk assessment of dynamic data changes, the first embodiment of the present invention provides a typhoon disaster chain risk assessment method, including but not limited to steps S101 to S102;
[0053] Step S101, obtaining first typhoon-related data at the current moment, and performing feature engineering on the first typhoon-related data to obtain multidimensional features;
[0054] In some embodiments, the first typhoon-related data includes typhoon meteorological data, urban disaster resistance data and disaster resistance measures data, wherein the typhoon meteorological data includes but is not limited to typhoon meteorological element data such as wind speed, wind direction, air pressure, temperature, humidity, and typhoon location, intensity, cloud distribution and other information; the urban disaster resistance data includes but is not limited to basic information of the city (such as population density, economic development level, infrastructure status, etc.), urban land use planning and emergency resource reserves, etc.; the disaster resistance measures data includes but is not limited to typhoon emergency plans, allocation of emergency resources and implementation progress of evacuation plans, etc.
[0055] In some embodiments, the first typhoon-related data at the current moment is obtained, specifically: the typhoon meteorological data can be obtained through, but is not limited to, weather stations, satellite remote sensing, and radar monitoring, the city's disaster resistance data is obtained through public information, and relevant disaster resistance measures data are collected through government public data, historical disaster management records, etc. The specific acquisition method is not the focus of this application, so it will not be expanded here.
[0056] In some embodiments, the feature engineering of the first typhoon-related data to obtain multidimensional features includes: performing data cleaning on the typhoon-related data to obtain a first processing result; normalizing the first processing result to obtain a second processing result; and performing feature extraction and feature construction on the second processing result to obtain multidimensional features. Specifically, first, the first typhoon-related data is cleaned and outliers are removed. For missing values such as air pressure, wind speed and precipitation in the first typhoon-related data, interpolation or data filling methods are used to fill in the missing values to ensure the integrity of all feature data, and the first processing result is obtained; secondly, due to the inconsistent measurement standards of different data sources, the cleaned first processing result needs to be normalized or standardized (for example, for precipitation data, we standardize the precipitation in different time periods according to the average precipitation of the city) so that data of different dimensions have the same dimension and magnitude, and the second processing result is obtained; finally, useful features (including threat factors, vulnerability factors and control measure factors, etc.) are extracted from the normalized second processing result, and new interactive features are constructed (such as "interactive features of precipitation and drainage capacity" to reflect the impact of heavy precipitation and the adaptability of urban drainage systems on flood risks) to enhance the expression ability of the model, and finally the above two features are combined to obtain multidimensional features.
[0057] It should be noted that threat factors describe the typhoon's intensity, path, precipitation, maximum wind speed, landing location, etc., which directly affect the destructive power and coverage of the disaster chain, and describe the natural characteristics of the typhoon itself and its direct impact; vulnerability factors reflect the disaster-bearing capacity of the disaster area, such as population density, economic development level, infrastructure disaster resistance, and accessibility of medical resources, and can reveal the differences in disaster tolerance in different regions; control measures reflect the role of human intervention in risk mitigation, such as early warning system coverage, resource allocation efficiency, and evacuation capacity.
[0058] It should be noted that for threat factors, since the threat of typhoons not only depends on a single meteorological parameter, but is also related to other external factors such as sea surface temperature and air pressure, we integrate multi-source data (including meteorological data, satellite remote sensing data, etc.) to obtain a more comprehensive and accurate threat factor.
[0059] It's important to note that changes in population density directly impact the difficulty of evacuating people during a disaster. Therefore, data normalization is necessary to make it suitable for further analysis. Different socioeconomic factors should also be combined to form new features that better capture regional vulnerability. Furthermore, by combining the "infrastructure condition" and "evacuation capacity" features, interactive features are generated to quantify a region's resilience to disasters.
[0060] In this way, by performing feature engineering on the first typhoon-related data, the first typhoon-related data can be converted into multi-dimensional features that can effectively characterize the risk of the typhoon disaster chain. The most valuable information for risk assessment can be quickly extracted from the first typhoon-related data, facilitating subsequent risk assessment, thereby improving the adaptability and predictive ability of the model.
[0061] Step S102: input the multidimensional features into a preset evaluation model, determine the key risk factors in the multidimensional features based on the feature importance weights, perform weighted prediction on the key risk factors through each decision tree, obtain the comprehensive prediction value of multiple decision trees, and determine the risk assessment level based on the comprehensive prediction value, wherein the feature importance weights are determined during the training process of the evaluation model, and each of the decision trees is obtained by residual fitting in the process of training the evaluation model using historical typhoon data, and the evaluation model is determined based on the model parameters optimized according to the prediction values of each decision tree.
[0062] In some embodiments, the training process of the evaluation model is specifically as follows: obtaining the historical typhoon data of a preset time, and performing feature engineering on the historical typhoon data to obtain a training data set; constructing an initial evaluation model, inputting the training data set into the initial evaluation model, and in each iteration, calculating the pseudo residual based on the current prediction value of the initial evaluation model, and using the pseudo residual as the target variable to fit the decision tree, and adding the prediction value corresponding to the current decision tree to the initial evaluation model to optimize the first model parameters until the preset iteration termination condition is met, and determining the evaluation model based on the target model parameters. Specifically, first, historical typhoon data such as historical meteorological data, socioeconomic data, geographic information data, and typhoon disaster event data within a preset time range are collected, and feature engineering is performed on the collected historical typhoon data, including steps such as data cleaning, normalization, feature extraction, and feature construction, to obtain a training data set for training. Secondly, a simple model is initialized through the training data set to assign the initial risk score y to all samples. i Set as the mean of the target variable (true value) as the initial prediction value F0(x), which is generally obtained by minimizing the loss function. The calculation formula of the initial prediction value F0(x) is Where F0(x) is the initial prediction value, and a constant value (such as the mean of the target variable or a constant value that minimizes the loss) is initialized as the initial prediction value for all samples; y i is the true value of the i-th sample, representing the true risk level of a scenario in the typhoon disaster chain risk assessment; n is the total number of training samples; γ is the initial constant value, which is the initial global value predicted by the model. All samples are assigned this value at this stage. Then, in each round of iteration, the prediction error (pseudo-residual) of the current model is calculated. The calculation formula for the pseudo-residual is: Where r im is the pseudo residual of the i-th sample in the m-th iteration, L(y i ,F(x i )) is the loss function, which is used to measure the model prediction value F(x i ) and the true value y i Then, using the pseudo residual r im A new decision tree is fitted with the target variable. Finally, after each iteration, the model updates its predictions and adds the predictions of the newly trained regression tree to the existing model's predictions. The prediction accuracy is further improved by optimizing the weights of each tree (the weights are determined by minimizing the loss function). This process continues until the preset number of iterations or the convergence condition of the loss function is reached. At this point, the evaluation model is determined based on the model parameters.
[0063] In some embodiments, the calculation formula for adding the predicted value of the current decision tree to the initial evaluation model to optimize the first model parameter is specifically:
[0064] F m (x) = F m-1 (x)+γ m h m (x);
[0065]
[0066] Where, F m (x) and F m-1 (x) are the overall predicted values for the mth and m-1th iterations, γ is the overall model parameter, and γ m is the model parameter determined in the mth time, h m (x) is the predicted value corresponding to the decision tree fitted at the mth iteration, y i is the true value of the i-th sample, x i The training data of the i-th sample, n is the total number of training samples.
[0067] In some embodiments, the true value y i It is an important input for model training and comes from the following aspects: (1) obtained through expert evaluation data, which is scored based on the expert experience of historical typhoon disaster chain events. These data provide the model with a real label of risk level; (2) from actual observation data, such as information on property losses, casualties, etc. recorded in typhoon disasters. These monitoring records are converted into quantitative risk scores through analysis; (3) statistical analysis data can be used to generate representative risk scores as the basis for evaluation based on the summary and extraction of historical disaster event data.
[0068] It should be noted that data on all typhoons experienced by the city over a preset time period (e.g., the past ten or five years) was collected through government meteorological departments and historical disaster records. This data includes meteorological parameters such as the typhoon's path, maximum wind speed, precipitation, air pressure, and path changes. Furthermore, socioeconomic data on the city was collected, including information such as population density, infrastructure development, urban disaster prevention capabilities, and accessibility to medical resources. The city's geographic information (such as topography, terrain, and drainage capacity) was also taken into consideration, as geographical features play a crucial role in secondary disasters such as floods and landslides.
[0069] It's important to note that the pseudo-residual is essentially the negative gradient of the loss function with respect to the current model's predictions, and can therefore be understood as guiding the direction of model optimization. Generally speaking, the pseudo-residual is calculated dynamically and adjusted in each iteration based on the latest model's predictions, gradually improving the model's prediction accuracy. By gradually fitting the pseudo-residual, prediction accuracy can be continuously optimized.
[0070] It's important to note that each splitting node in a decision tree is used to capture the impact of input features on pseudo-residuals and reveal the importance of key factors. The decision tree selects the feature that most reduces prediction error for splitting, and through continuous splitting, it gradually optimizes the model's prediction accuracy. For example, in this typhoon disaster chain assessment, it may be found that maximum wind speed and precipitation are the two most important features affecting disaster severity. These two features may become the most frequently split nodes in the decision tree, thereby determining which factors contribute most to the pseudo-residuals.
[0071] It should be noted that the evaluation model can be but is not limited to GBDT, which is also known as Gradient Boosting Decision Tree. As an efficient integrated learning algorithm, GBDT has shown great potential in typhoon disaster chain risk analysis. GBDT combines the advantages of nonlinear modeling, strong robustness, dynamic update capabilities and feature importance analysis, and can effectively deal with complex nonlinear relationships and multidimensional variable characteristics in disaster chains. In addition, GBDT has a strong tolerance for missing data and noisy data, and is suitable for scenarios with incomplete disaster data. In terms of dynamic updates, GBDT can adjust prediction results based on real-time data, such as quickly outputting new risk assessments when typhoon paths and intensities change. At the same time, GBDT can help identify the most critical risk factors in disaster chains through the feature importance ranking function, providing decision makers with a clear basis for action.
[0072] It's important to note that after training the evaluation model, its performance needs to be evaluated. This can be achieved through cross-validation, which involves dividing the data into multiple subsets, using one subset as the validation set and the remaining subsets as the training set to evaluate the model's performance. This approach effectively avoids overfitting and ensures good generalization. For typhoon disaster chain risk assessment, cross-validation can help test the prediction stability of the evaluation model across different typhoon events and adjust the model's hyperparameters based on the different validation results.
[0073] In this way, by training the evaluation model, it is convenient for subsequent evaluation models to complete the processing and prediction of large amounts of data in a shorter period of time, meeting the needs of rapid response under dynamic data changes, while effectively capturing the complex interactions and nonlinear relationships between various factors in the typhoon disaster chain, and significantly improving the prediction accuracy of risk assessment.
[0074] In some embodiments, the multidimensional features are input into a preset assessment model, and the key risk factors in the multidimensional features are determined based on the feature importance weights. Specifically, the feature importance weights determined by the assessment model during training are used to screen out the key risk factors that have the greatest impact on the risk assessment results (which features have the greatest impact on the risk assessment results) from the multidimensional features, where the feature importance weights reflect the contribution of each feature to the model's prediction results. This method can effectively select those threat factors (such as wind speed, precipitation, etc.), vulnerability factors, and control measures factors that can truly affect the risk of the disaster chain.
[0075] It should be noted that the above method ensures that only core factors reflecting the post-disaster recovery capacity and evacuation capacity of the affected area are selected for the final model. At the same time, considering that post-disaster recovery capacity and vulnerability factors are dynamic, a time factor needs to be added to address the impact of factors such as urbanization and population mobility on vulnerability factors.
[0076] It's important to note that by incorporating real-time data feedback mechanisms and adjusting the assessment weights of control measures, risk assessments can be made more dynamically adaptable. By summarizing historical typhoon disaster management experience, evaluating the effectiveness of existing control measures, and using models to simulate the impact of strengthening different control measures on risk assessment results, such as enhancing post-disaster recovery capabilities and expanding the coverage of early warning systems, the risk level can be significantly reduced. For example, if the model indicates that precipitation is a key factor, emergency management departments can prioritize areas with high precipitation and deploy drainage systems and flood control facilities in advance.
[0077] By selecting and optimizing threat, vulnerability, and control factors, this approach combines multi-dimensional data to accurately identify key factors critical to typhoon disaster chain risk assessment, reducing biases in subjective judgment. This not only improves assessment accuracy but also offers robust dynamic updating and adaptability, providing scientific and comprehensive support for disaster prevention and mitigation decision-making.
[0078] In some embodiments, the weighted prediction of the key risk factors by each decision tree to obtain a comprehensive prediction value of multiple decision trees includes: for each decision tree, judging from the root node to the leaf node according to the key risk factors, and outputting a number of initial prediction values; weighted summing up each of the initial prediction values according to the model weight to obtain a comprehensive prediction value of multiple decision trees. Specifically, there are multiple decision trees in the evaluation model. When determining the key risk factors, each decision tree will judge from the root node to the leaf node according to the input risk factors to give the corresponding initial prediction value. After that, the initial prediction values are weighted summed up according to the weight of each decision tree in the evaluation model (determined during training), and the comprehensive prediction value of multiple decision trees, that is, the evaluation model, can be obtained.
[0079] In this way, by performing weighted prediction on the key risk factors through each decision tree, the prediction can be completed in a shorter time, ensuring the efficiency of the entire evaluation process. At the same time, the weighted prediction integrates the results of multiple decision trees, which can better fit the actual distribution of the data and improve the accuracy of the prediction.
[0080] It's important to note that the assessment model's output isn't just a comprehensive forecast; it can also enhance decision-making support through scenario simulations. By simulating the impact of different typhoon paths, intensities, or disaster prevention measures on the disaster chain risk, the model can help emergency management departments evaluate the effectiveness of different response strategies. For example, by increasing the coverage of early warning systems, strengthening post-disaster recovery capabilities, or optimizing evacuation procedures, the model can predict the impact of these measures on overall disaster risk mitigation. These scenario simulations provide decision makers with a clear operational basis, helping them select the most effective emergency response plan.
[0081] In some embodiments, the risk assessment level is determined based on the comprehensive prediction value. Specifically, after the comprehensive prediction value is determined, it is necessary to convert the comprehensive prediction value into a risk assessment level according to a preset threshold, such as four levels of low risk, medium risk, high risk and extremely high risk. This form of result output facilitates decision makers to quickly judge the urgency of the disaster chain and give priority to formulating corresponding prevention and control strategies.
[0082] In some embodiments, after the risk assessment level is determined based on the comprehensive prediction value, it also includes: obtaining the second typhoon-related data at the current moment; inputting the second typhoon-related data into the evaluation model, adjusting the target model parameters of the evaluation model through an incremental learning method to obtain second model parameters; and determining the updated risk assessment level based on the second model parameters. Specifically, after the risk assessment level is evaluated, when relevant data such as the typhoon path and intensity change, the second typhoon-related data at the current moment is obtained, and the second typhoon-related data is input into the evaluation model after data cleaning and feature engineering; then, the pseudo residual is calculated based on the predicted value and the actual value of the current model (understood as the comprehensive prediction value of the previous stage), and the pseudo residual is used as the target variable to fit a new decision tree, and the predicted value of the new decision tree is added to the overall evaluation model, and the model performance is further improved by optimizing the weights to obtain the second model parameters; finally, the multi-dimensional features of the input are predicted based on the determined second model parameters to obtain a new comprehensive prediction value, and the risk assessment level is updated based on the new comprehensive prediction value.
[0083] It's important to note that through incremental learning, the model can process new data as it arrives, without the need to retrain the entire model. With real-time data, the model can adjust disaster risk assessments based on the latest typhoon developments and adjust resource allocation and emergency response measures based on the changing risk level. For example, when a typhoon's path shifts, or when precipitation and wind speed change dramatically, the assessment model can quickly update risk scores and provide real-time decision support to emergency management departments. If the model indicates a sharp increase in risk levels in certain areas after an update, the government can immediately initiate evacuation plans and mobilize additional resources for rescue efforts. If the risk level in certain areas decreases, manpower and material resources can be reduced and relocated to higher-risk areas.
[0084] In this way, the target model parameters of the evaluation model are adjusted by the incremental learning method, which can
[0085] Ensure that the model can adjust the disaster risk assessment level according to the latest dynamics of the typhoon.
[0086] In some embodiments, after the risk assessment level is determined based on the comprehensive prediction value, it also includes: determining the corresponding first disaster prevention strategy based on the risk assessment level according to a preset first matching rule; determining the corresponding second disaster prevention strategy based on the key risk factor according to a preset second matching rule; determining the target disaster prevention strategy based on the first disaster prevention strategy and the second disaster prevention strategy, and performing disaster prevention according to the target disaster prevention strategy. Specifically, first, after determining the risk assessment level, since disaster prevention strategies corresponding to different risk levels have been set in advance based on historical experience and expert knowledge, it is only necessary to directly match the corresponding first disaster prevention strategy based on the risk level; secondly, since disaster prevention strategies corresponding to different key risk factors have been set in advance based on historical experience and expert knowledge, it is only necessary to directly match the corresponding second disaster prevention strategy based on the key risk factor; then, by comprehensively considering the impact of the risk assessment level and risk factors, the target disaster prevention strategy can be determined, thereby ensuring the comprehensiveness and pertinence of the disaster prevention measures; finally, according to the target disaster prevention strategy, specific disaster prevention and mitigation measures are organized and implemented.
[0087] It should be noted that the first matching rule can be: (1) Low risk level: take basic monitoring and early warning measures, appropriately reserve emergency supplies and equipment, and carry out disaster prevention knowledge publicity; (2) Medium risk level: strengthen monitoring and early warning, pre-position emergency supplies and equipment, organize emergency rescue teams to be on standby, and formulate detailed personnel evacuation plans. (3) High risk level: comprehensive monitoring and early warning, comprehensive deployment of emergency supplies, equipment and rescue teams, take key area protection measures, initiate personnel evacuation plans, and implement traffic control. (4) Extremely high risk level: declare a state of emergency, forcibly evacuate all residents in potentially affected areas, take special protection measures, and organize multi-department joint operations.
[0088] It should be noted that the second matching rule can be: (1) Threat factors: If wind speed and precipitation are the main threat factors, strengthen meteorological monitoring and forecasting, and take engineering protection measures such as strengthening buildings and flood control facilities. (2) Vulnerability factors: If population density and infrastructure conditions are the main vulnerability factors, optimize population evacuation and resettlement planning and strengthen infrastructure construction. (3) Control measures factors: If early warning system coverage and resource allocation efficiency are the key control measures, optimize the early warning system and improve the efficiency of emergency resource reserves and management.
[0089] It should be noted that disaster prevention and mitigation measures can be: (1) Comprehensive protection measures: Combining meteorological monitoring, engineering protection, infrastructure reinforcement, early warning system optimization and other measures to form a comprehensive disaster prevention and mitigation plan. (2) Resource allocation: Rationally allocate resources according to the target disaster prevention strategy to ensure sufficient resource support in key areas and key links. (3) Emergency response: Guide emergency management departments to formulate and implement specific emergency response plans to improve the ability and efficiency of disaster response.
[0090] In this way, by determining the target disaster prevention strategy through risk assessment levels and key risk factors, we can ensure that disaster prevention measures are both comprehensive and targeted, and improve the scientific nature and rationality of disaster prevention strategies.
[0091] As shown in the typhoon disaster chain risk assessment method, based on the above method embodiment, a corresponding device embodiment is provided, please refer to Figure 2 , comprising: an acquisition module 100 and an evaluation module 200;
[0092] The acquisition module is used to acquire the first typhoon-related data at the current moment, and perform feature engineering on the first typhoon-related data to obtain multidimensional features;
[0093] The evaluation module is used to input the multidimensional features into a preset evaluation model, determine the key risk factors in the multidimensional features based on the feature importance weights, and perform weighted predictions on the key risk factors through each decision tree to obtain a comprehensive prediction value of multiple decision trees, and determine the risk assessment level based on the comprehensive prediction value, wherein the feature importance weights are determined during the training process of the evaluation model, and each of the decision trees is obtained by residual fitting in the process of training the evaluation model using historical typhoon data. The evaluation model is determined based on the model parameters optimized according to the prediction values of each decision tree.
[0094] It can be understood that the above-mentioned device embodiment corresponds to the method embodiment of the present invention, and it can implement any one of the above-mentioned method embodiments of the present invention to provide a typhoon disaster chain risk assessment method.
[0095] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. Furthermore, in the drawings of the device embodiments provided by the present invention, the connection relationship between modules indicates that they have a communication connection, which may be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement the present invention without inventive effort.
[0096] Based on the above-mentioned embodiment of the typhoon disaster chain risk assessment method, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the typhoon disaster chain risk assessment method of any embodiment of the present invention is implemented.
[0097] For example, in this embodiment, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the terminal device.
[0098] The terminal device may be a computing device such as a desktop computer, a notebook computer, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.
[0099] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, connecting various parts of the entire terminal device using various interfaces and lines.
[0100] Based on the above-mentioned method embodiments, another embodiment of the present invention provides a computer-readable storage medium, including a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the typhoon disaster chain risk assessment method described in any one of the above-mentioned method embodiments of the present invention.
[0101] Wherein, the module / unit integrated in the device / terminal equipment, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc.
[0102] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.
Claims
1. A typhoon disaster chain risk assessment method, characterized in that: include: Acquire first typhoon-related data at the current moment, and perform feature engineering on the first typhoon-related data to obtain multidimensional features; The multidimensional features are input into a preset evaluation model, and the key risk factors in the multidimensional features are determined based on the feature importance weights. The key risk factors are weightedly predicted through each decision tree to obtain a comprehensive prediction value of multiple decision trees, and the risk assessment level is determined based on the comprehensive prediction value, wherein the feature importance weights are determined during the training process of the evaluation model, and each of the decision trees is obtained by residual fitting in the process of training the evaluation model using historical typhoon data. The evaluation model is determined based on the model parameters optimized according to the prediction values of each decision tree.
2. The typhoon disaster chain risk assessment method according to claim 1, characterized in that: The feature engineering is performed on the first typhoon-related data to obtain multi-dimensional features, specifically: Performing data cleaning on the typhoon-related data to obtain a first processing result; Normalizing the first processing result to obtain a second processing result; Feature extraction and feature construction are performed on the second processing result to obtain multi-dimensional features.
3. The typhoon disaster chain risk assessment method according to claim 1, characterized in that: The training process of the evaluation model is specifically as follows: Acquire the historical typhoon data of a preset time, and perform feature engineering on the historical typhoon data to obtain a training data set; An initial evaluation model is constructed, and the training data set is input into the initial evaluation model. In each iteration, a pseudo residual is calculated based on the current prediction value of the initial evaluation model, and a decision tree is fitted using the pseudo residual as the target variable. The prediction value corresponding to the current decision tree is added to the initial evaluation model to optimize the first model parameters until a preset iteration termination condition is met, and an evaluation model is determined based on the target model parameters.
4. The typhoon disaster chain risk assessment method according to claim 3, characterized in that: The calculation formula for adding the predicted value of the current decision tree to the initial evaluation model to optimize the first model parameter is specifically: F m (x)=F m-1 (x)+γ m h m (x); Where, F m (x) and F m-1 (x) are the overall predicted values for the mth and m-1th iterations, γ is the overall model parameter, and γ m is the model parameter determined in the mth time, h m (x) is the predicted value corresponding to the decision tree fitted at the mth iteration, y i is the true value of the i-th sample, x i The training data of the i-th sample, n is the total number of training samples.
5. The typhoon disaster chain risk assessment method according to claim 1, characterized in that: The key risk factors are weighted predicted by each decision tree to obtain a comprehensive prediction value of multiple decision trees, specifically: For each decision tree, make judgments from the root node to the leaf node based on the key risk factors and output several initial prediction values; The initial prediction values are weighted and summed according to the model weights to obtain the comprehensive prediction value of multiple decision trees.
6. The typhoon disaster chain risk assessment method according to any one of claims 1 to 5, characterized in that: After determining the risk assessment level based on the comprehensive prediction value, the method further includes: Get the second typhoon related data at the current moment; Inputting the second typhoon-related data into the evaluation model, and adjusting the target model parameters of the evaluation model through an incremental learning method to obtain second model parameters; Based on the second model parameters, an updated risk assessment level is determined.
7. The typhoon disaster chain risk assessment method according to any one of claims 1 to 5, characterized in that: After determining the risk assessment level based on the comprehensive prediction value, the method further includes: Determining a corresponding first disaster prevention strategy based on the risk assessment level according to a preset first matching rule; Determining a corresponding second disaster prevention strategy based on the key risk factors according to a preset second matching rule; A target disaster prevention strategy is determined based on the first disaster prevention strategy and the second disaster prevention strategy, and disaster prevention is performed according to the target disaster prevention strategy.
8. A typhoon disaster chain risk assessment system, characterized in that: include: Access modules and assessment modules; The acquisition module is used to acquire the first typhoon-related data at the current moment, and perform feature engineering on the first typhoon-related data to obtain multidimensional features; The evaluation module is used to input the multidimensional features into a preset evaluation model, determine the key risk factors in the multidimensional features based on the feature importance weights, and perform weighted predictions on the key risk factors through each decision tree to obtain a comprehensive prediction value of multiple decision trees, and determine the risk assessment level based on the comprehensive prediction value, wherein the feature importance weights are determined during the training process of the evaluation model, and each of the decision trees is obtained by residual fitting in the process of training the evaluation model using historical typhoon data. The evaluation model is determined based on the model parameters optimized according to the prediction values of each decision tree.
9. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the method for assessing the risk of a typhoon disaster chain according to any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium, characterized in that include: A stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the typhoon disaster chain risk assessment method according to claims 1-7.