Rainstorm disaster loss assessment method and system based on machine learning

Abstract

The present invention discloses a method and system for rainstorm disaster loss assessment based on machine learning, comprising the following steps: taking data of a rainstorm disaster data source in a preset area within a specified time period as data to be analyzed; dividing the preset area into regions based on rainfall according to the data to be analyzed, taking the preset area with a rainfall greater than a threshold as a disaster area, and taking the preset area with a rainfall greater than a threshold as an uncertain area; calculating the destructive power of the uncertain area according to the hydrological mechanism and the drainage system, and adding the uncertain area with a rainfall greater than the threshold to the disaster area; performing a comprehensive risk analysis on the data to be analyzed in the disaster area to obtain a comprehensive risk index, calculating a damage degree index of the disaster area, and performing a real-time dynamic assessment of the disaster area according to the damage degree index and the comprehensive risk index to obtain a loss index; constructing a rainstorm disaster loss assessment model according to the loss index, and outputting an assessment result.

Description

Technical Field

[0001] The present invention relates to the field of disaster loss assessment, and in particular to a method and system for rainstorm disaster loss assessment based on machine learning. Background Art

[0002] With global climate change and accelerating urbanization, the frequent occurrence of rainstorm disasters has become a major challenge threatening urban safety and socioeconomic stability. Traditional disaster assessment methods rely heavily on historical disaster statistics and empirical models, suffering from drawbacks such as a single data dimension and delayed dynamic response. Regional demarcation methods based on rainfall intensity thresholds often overlook the spatial heterogeneity of the hydrological cycle and urban drainage systems, resulting in inaccurate identification of disaster scope. Furthermore, existing research has largely focused on physical damage indicators and has failed to effectively integrate social perception data such as real-time social media sentiment and regional economic vulnerability, resulting in significant deviations between loss assessment results and post-disaster recovery needs.

[0003] In recent years, machine learning technology has achieved breakthroughs in disaster prediction. The application of convolutional neural networks in remote sensing imagery for damage identification, and the LSTM model's ability to process hydrological time series data, have significantly improved disaster monitoring efficiency. However, existing models generally face two major bottlenecks: the lack of physical constraints in the fusion of features from multi-source heterogeneous data, resulting in reduced model interpretability; and insufficient generalization across disaster scenarios in different regions, requiring repeated training with large amounts of labeled data. However, by constructing shared feature representations across tasks, the meta-learning framework demonstrates unique transfer advantages in small sample size scenarios, providing new insights for building cross-regional, multi-scenario rainstorm disaster assessment models.

[0004] Current research gaps lie in three key areas: a dynamic coupling mechanism for multimodal data has yet to be established, particularly for modeling the spatiotemporal correlation between geological permeability and urban pipe network topology; dynamic quantification of socioeconomic vulnerability during the evolution of disaster chains requires improvement; and existing assessment models inefficiently reuse knowledge from historical disaster scenarios. Therefore, a new approach to stormwater loss assessment is urgently needed. By integrating hydrological mechanism models with a meta-learning paradigm, this approach can overcome the static assessment limitations of traditional methods and achieve multi-dimensional dynamic perception and cross-regional migration assessment of disaster losses. Summary of the Invention

[0005] The purpose of the present invention is to provide a rainstorm disaster loss assessment method based on machine learning.

[0006] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0007] The present invention comprises the following steps:

[0008] Data from a rainstorm disaster data source in a preset area within a specified time period is used as the data to be analyzed; the data source to be analyzed includes meteorological data, geological data, topographic data, socioeconomic data, social media data, historical disaster data, ecological data, and architectural data;

[0009] The method divides the preset area into regions based on the data to be analyzed based on rainfall, defines the preset area with a rainfall greater than a threshold as a disaster area, and defines the area with a rainfall greater than a threshold as an uncertain area, calculates the destructive force of the uncertain area based on the hydrological mechanism and the drainage system, and adds the uncertain area with a rainfall greater than the threshold to the disaster area; the method includes:

[0010] Calculate the building's exposure, sensitivity, and adaptability based on building data, and calculate the destructive power through objective weighting:

[0011]

[0012] BR i =ρ4WE i +ρ5WN i -ρ6WS i

[0013] The building vulnerability of the i-th uncertain area is BR i , the exposure of the i-th uncertainty area is WE i , the sensitivity of the i-th uncertainty region is WN i , the adaptability of the i-th uncertain region is WS i , the exposure weight is ρ4, the sensitivity weight is ρ5, the adaptability is ρ6, and the destructive power of the i-th uncertainty area is The rainfall in the i-th uncertain area is P i ;

[0014] Performing a comprehensive risk analysis on the data to be analyzed in the disaster area to obtain a comprehensive risk index, calculating a damage degree index of the disaster area, and performing a real-time dynamic assessment of the disaster area based on the damage degree index and the comprehensive risk index to obtain a loss index;

[0015] A rainstorm disaster loss assessment model based on meta-learning is constructed according to the loss index, the data to be assessed is input into the rainstorm disaster loss assessment model, and an assessment result is output.

[0016] Furthermore, the method for dividing the preset area according to the data to be analyzed based on rainfall includes:

[0017] The data to be analyzed and the preset area are aligned in time and space, and the expert method is used to perform primary division of the preset area to obtain the initial area. The rainfall in the initial area is calculated based on the data to be analyzed. The initial area with a rainfall greater than the threshold is regarded as a disaster area, otherwise it is regarded as an uncertain area.

[0018] Furthermore, the method for disaster areas includes:

[0019] The rainfall threshold is divided into ordinary rainfall threshold and super-threshold rainfall, and the disaster area is divided into ordinary disaster area and super-heavy rainfall disaster area;

[0020] Disaster areas with rainfall amounts above the threshold for normal rainfall but less than the threshold for heavy rainfall are classified as normal disaster areas; disaster areas with rainfall amounts above the threshold for heavy rainfall are classified as super heavy rainfall disaster areas.

[0021] Normal rainfall thresholds and above-threshold rainfall are dynamically adjusted according to the rainfall threshold drift due to climate warming.

[0022] Furthermore, the method for performing comprehensive risk analysis on the data to be analyzed in the disaster area includes:

[0023] The disaster-inducing factor index of common disaster areas is calculated through meteorological data, the geological vulnerability index of disaster areas is calculated through geological data, the terrain stability index is calculated through topographic and geomorphological data, and the building vulnerability is calculated through building data;

[0024] Ecological sensitivity, ecological resilience, and ecological pressure are calculated through ecological data, and the ecological vulnerability index is calculated based on ecological sensitivity, ecological resilience, and ecological pressure:

[0025]

[0026] The ecological vulnerability index of the u-th common disaster area is Ecological sensitivity is C1, ecological pressure is C2, ecological resilience is C3, sensitivity weight is ρ1, pressure weight is ρ2, and resilience weight is ρ3;

[0027] Use disaster data and social media data to revise the disaster inducing factor index, geological vulnerability index, and terrain stability index;

[0028] The comprehensive risk index of common rainstorm disaster areas is calculated based on the disaster inducing factor index, geological vulnerability index, terrain stability index and ecological vulnerability index:

[0029]

[0030] The comprehensive risk index of the u-th common disaster area is The disaster inducing factor index of the u-th common disaster area is MHu , the geological vulnerability index of the u-th general disaster area is DX u , the terrain stability index of the u-th common disaster area is ZX u , the induced weight is ω1, the geological weight is ω2, the topographic weight is ω3, and the ecological weight is The building weight is ω5, and the building vulnerability of the u-th common disaster area is BR u ;

[0031] Calculate the comprehensive risk index of superstorm disaster areas:

[0032]

[0033] The comprehensive risk index of the u-th super rainstorm disaster area is The disaster area is marked according to the comprehensive risk index and the marked disaster area is output.

[0034] Furthermore, the method for calculating the damage degree index of the disaster area includes:

[0035] Calculate the loss of life in the disaster area based on the data to be analyzed:

[0036]

[0037] The loss of life is LM s , the number of disaster areas is N a , the number of risk population is N p , the mortality rate of the risk population is η, and the influence of the bth major influencing factor on the mortality rate of the risk population is γ b , the weight coefficient of the bth main influencing factor is φ b , the influence of the bth secondary factor on the mortality rate of the risk population is ξ b , the weight coefficient of the bth secondary factor is The number of secondary factors is N b2 , the number of main factors is N b1 , the weight coefficient is β, and the correction coefficient is

[0038] Calculate the economic losses in the disaster area:

[0039]

[0040] The economic loss is D s The number of pixel disaster areas where properties are distributed is N1, the type of properties in the disaster area of ​​the vth pixel is N2, the number of different flooding depth levels is N3, and the loss rate of the vth type of property in the ath disaster area corresponding to the xth flooding depth is The property value of the vth type of property in the ath disaster area corresponding to the xth flooding depth is

[0041] Calculate the indirect economic losses in the disaster area:

[0042]

[0043] The indirect coefficient corresponding to the rth type of property is K a (s), the number of property categories is M1, and the rth property is A r , the indirect economic loss is

[0044] Calculate the ecological environment loss rate:

[0045]

[0046] The ecological environment loss rate is E, and the loss rate of the c-type ecosystem service function is Q c , the number of ecosystem service functions is N4, the number of ecological products is N5, and the service volume of the i-th product in the c-th ecosystem service function is H ck , the unit price of the i-th product in the c-th ecosystem service function is R ck ;

[0047] The damage index is calculated based on the loss of life, economic losses, indirect economic losses, and ecological environment loss rates:

[0048]

[0049] The total economic volume is KS, and the damage index is

[0050] Furthermore, the method of performing a real-time dynamic assessment of the disaster area according to the damage degree index and the comprehensive risk index to obtain a loss index includes:

[0051] The damage degree index and comprehensive risk index are input into a real-time dynamic assessment model based on a Bayesian network, and the state-time curve is used to discretize the disaster state into three levels: extremely heavy rain, heavy rain, and torrential rain.

[0052] Taking time as a node, calculate the node's transition network conditional probability:

[0053]

[0054] The time interval is Δt, and the offset rate when node d changes from state w to state m is The state of the node at time t0 is w, and the state of the node at time t1 is j. Under the condition that the state is w at time t0, the probability of the node being in state j at time t1 is k(G t1 =j|G t0 =w), the damage index at time t0 is The comprehensive risk index at time t0 is The damage index of node d in state w is The damage index of node d in state m is The comprehensive risk index of node d in state w is The comprehensive risk index of node d in state m is

[0055] Calculate the loss index based on the conditional probability of the transfer network:

[0056]

[0057] The damage index of state j at time t is The comprehensive risk index in state j at time t is The loss index at time t is The damage index at time t in state w is The comprehensive risk index in state w at time t is

[0058] Furthermore, a method for constructing a rainstorm disaster loss assessment model based on meta-learning according to the loss index includes:

[0059] The mean square error between actual rainstorm disaster loss assessment and predicted disaster loss is used as the assessment loss function, and the weighted addition of loss index and assessment loss function is used as the objective function of the rainstorm disaster loss assessment model.

[0060] Initialize the parameters of the rainstorm disaster loss assessment model and use the long short-term memory network as the predictor of the rainstorm disaster loss assessment model;

[0061] Randomly sample the data to be analyzed from the meta-training task, use the support set to calculate the inner gradient, and update the parameters of the rainstorm disaster loss assessment model. The expression is:

[0062]

[0063] The qth meta-training task is T q , the inner learning rate is α, and the initial parameters are Meta-training task T q The updated parameters of the inner layer are Meta-training task T q The loss function is about the initial parameters The gradient of Initial parameters The rainstorm disaster loss assessment model is

[0064] Compute the meta-loss using the support set:

[0065]

[0066] The loss of the The number of meta-training tasks is Using parameters on querysets Compute the meta-training task T q The loss is Update the outer parameters, the expression is:

[0067]

[0068]

[0069] The outer layer learning rate is ψ, and the updated outer layer parameters are parameter The rainstorm disaster loss assessment model is Meta-loss with respect to parameters The gradient of

[0070] Output the rainstorm disaster loss assessment model after adding meta-learning.

[0071] The second aspect is the rainstorm disaster loss assessment system based on machine learning, which includes:

[0072] Data collection module: used to collect data from rainstorm disaster data sources in a preset area within a specified time period as data to be analyzed; the data sources to be analyzed include meteorological data, geological data, socioeconomic data, disaster data, social media data, and historical disaster data; the historical disaster data includes historical meteorological data, historical geological data, historical socioeconomic data, and historical disaster data;

[0073] Disaster area division module: used to divide the preset area into regions based on rainfall according to the data to be analyzed, and to divide the preset area with rainfall greater than a threshold into disaster areas, and vice versa as uncertain areas, and to calculate the destructive force of the uncertain area based on the hydrological mechanism and drainage system, and to add the uncertain area with rainfall greater than the threshold into the disaster area;

[0074] An analysis and real-time dynamic evaluation module is configured to perform a comprehensive risk analysis on the data to be analyzed in the disaster area to obtain a comprehensive risk index, calculate a damage degree index of the disaster area, and perform a real-time dynamic evaluation of the disaster area based on the damage degree index and the comprehensive risk index to obtain a loss index;

[0075] Modeling output module: used to construct a rainstorm disaster loss assessment model based on the loss index, input the data to be assessed into the rainstorm disaster loss assessment model, and output the assessment results.

[0076] The beneficial effects of the present invention are:

[0077] The present invention is a method and system for rainstorm disaster loss assessment based on machine learning. Compared with the existing technology, the present invention has the following technical effects:

[0078] The present invention can improve the accuracy of machine learning-based rainstorm disaster loss assessment through regional division, obtaining disaster areas, calculating the degree of destruction index, real-time dynamic evaluation and model construction steps, thereby improving the precision of machine learning-based rainstorm disaster loss assessment, optimizing machine learning-based rainstorm disaster loss assessment, greatly saving resources, and improving work efficiency. It can realize intelligent assessment of machine learning-based rainstorm disaster losses, perform real-time regional division and real-time dynamic evaluation of machine learning-based rainstorm disaster loss assessment, which is of great significance to machine learning-based rainstorm disaster loss assessment, can adapt to machine learning-based rainstorm disaster loss assessment of different standards and different machine learning-based rainstorm disaster loss assessment needs, and has a certain universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 This is a flowchart of the steps of the rainstorm disaster loss assessment method based on machine learning of the present invention. DETAILED DESCRIPTION

[0080] The present invention will be further described below through specific examples. The illustrative examples and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0081] The rainstorm disaster loss assessment method and system based on machine learning of the present invention include the following steps:

[0082] like Figure 1 As shown, in this embodiment, the following steps are included:

[0083] Data from a rainstorm disaster data source in a preset area within a specified time period is used as the data to be analyzed; the data source to be analyzed includes meteorological data, geological data, topographic data, socioeconomic data, social media data, historical disaster data, ecological data, and architectural data;

[0084] In the actual assessment, the residential area of ​​a certain city was selected as the research object, and data for the heavy rain in late summer of 2023 were collected for analysis. The meteorological data included 35 mm of rainfall in the past 12 hours, 55 mm of rainfall in the past 24 hours, a wind speed of 15 m / s, and a southeasterly wind direction. The geological data included clay soil, a groundwater level of 2 m, and a stable geological structure. The topographic data included an altitude of 50 m, a slope of 5 degrees, and a plain terrain. The socioeconomic data included a population density of 1,000 people / km2, an average income of 50,000 yuan / year per capita, a secondary industry share of 40%, and a tertiary industry share of 60%. The social media data included a high level of discussion about the heavy rain in the past 12 hours, and 5 likes on related topics. 000, forwarding number 2000; historical disaster data includes 1 small-scale waterlogging incident in the past week, 0 flood disasters in the past month, 3 rainstorm disasters in the region in the past 5 years, and an average economic loss of 79.04 million yuan; ecological data includes a decrease in vegetation coverage from 57% to 32%, and a decrease in wetland area from 21% to 10%; building data includes 500 buildings, 309 residential buildings, 157 commercial buildings, and 44 industrial buildings, with a building exposure of 0.6, a building sensitivity of 0.53, a building adaptability of 0.409, an exposure weight of 0.41, a sensitivity weight of 0.297, and an adaptability weight of 0.3;

[0085] The method divides the preset area into regions based on the data to be analyzed based on rainfall, defines the preset area with a rainfall greater than a threshold as a disaster area, and defines the area with a rainfall greater than a threshold as an uncertain area, calculates the destructive force of the uncertain area based on the hydrological mechanism and the drainage system, and adds the uncertain area with a rainfall greater than the threshold to the disaster area; the method includes:

[0086] Calculate the building's exposure, sensitivity, and adaptability based on building data, and calculate the destructive power through objective weighting:

[0087]

[0088] BR i =ρ4WE i +ρ5WN i -ρ6WS i

[0089] The building vulnerability of the i-th uncertain area is BR i , the exposure of the i-th uncertainty area is WE i , the sensitivity of the i-th uncertainty region is WN i , the adaptability of the i-th uncertain region is WS i , the exposure weight is ρ4, the sensitivity weight is ρ5, the adaptability is ρ6, and the destructive power of the i-th uncertainty area is The rainfall in the i-th uncertain area is P i ;

[0090] In the actual assessment, the initial areas are A1, A2, A3, B1, B2, and B3, the disaster areas are A1, A2, and B2, and the uncertain areas are A3, B1, and B3; the uncertain area added to the disaster area is B3;

[0091] The rainfall threshold is 29 mm in 12 hours and 49 mm in 24 hours; the destructive power threshold is 7.12;

[0092] Performing a comprehensive risk analysis on the data to be analyzed in the disaster area to obtain a comprehensive risk index, calculating a damage degree index of the disaster area, and performing a real-time dynamic assessment of the disaster area based on the damage degree index and the comprehensive risk index to obtain a loss index;

[0093] In the actual assessment, the comprehensive risk indexes of A1, A2, B2, and B3 were 0.91, 0.63, 0.74, and 0.85 respectively; the damage degree indexes of A1, A2, B2, and B3 were 0.74, 0.69, 0.72, and 0.89 respectively; and the loss indexes of A1, A2, B2, and B3 were 0.73, 0.67, 0.75, and 0.903 respectively;

[0094] Constructing a rainstorm disaster loss assessment model based on meta-learning according to the loss index, inputting the data to be assessed into the rainstorm disaster loss assessment model, and outputting an assessment result;

[0095] In actual evaluation, the evaluation error of the rainstorm disaster loss assessment model was reduced by 29% after adding meta-learning.

[0096] In this embodiment, the method for dividing the preset area according to the data to be analyzed based on rainfall includes:

[0097] The data to be analyzed and the preset area are aligned in time and space, and the expert method is used to perform primary division of the preset area to obtain the initial area. The rainfall in the initial area is calculated based on the data to be analyzed. The initial area with a rainfall greater than the threshold is regarded as a disaster area, otherwise it is regarded as an uncertain area.

[0098] In this embodiment, the method for disaster area includes:

[0099] The rainfall threshold is divided into ordinary rainfall threshold and super-threshold rainfall, and the disaster area is divided into ordinary disaster area and super-heavy rainfall disaster area;

[0100] Disaster areas with rainfall amounts above the threshold for normal rainfall but less than the threshold for heavy rainfall are classified as normal disaster areas; disaster areas with rainfall amounts above the threshold for heavy rainfall are classified as super heavy rainfall disaster areas.

[0101] Normal rainfall thresholds and above-threshold rainfall are dynamically adjusted according to the rainfall threshold drift due to climate warming.

[0102] In this embodiment, the method for performing comprehensive risk analysis on the data to be analyzed in the disaster area includes:

[0103] The disaster-inducing factor index of common disaster areas is calculated through meteorological data, the geological vulnerability index of disaster areas is calculated through geological data, the terrain stability index is calculated through topographic and geomorphological data, and the building vulnerability is calculated through building data;

[0104] Ecological sensitivity, ecological resilience, and ecological pressure are calculated through ecological data, and the ecological vulnerability index is calculated based on ecological sensitivity, ecological resilience, and ecological pressure:

[0105]

[0106] The ecological vulnerability index of the u-th common disaster area is Ecological sensitivity is C1, ecological pressure is C2, ecological resilience is C3, sensitivity weight is ρ1, pressure weight is ρ2, and resilience weight is ρ3;

[0107] Use disaster data and social media data to revise the disaster inducing factor index, geological vulnerability index, and terrain stability index;

[0108] The comprehensive risk index of common rainstorm disaster areas is calculated based on the disaster inducing factor index, geological vulnerability index, terrain stability index and ecological vulnerability index:

[0109]

[0110] The comprehensive risk index of the u-th common disaster area is The disaster inducing factor index of the u-th common disaster area is MH u , the geological vulnerability index of the u-th general disaster area is DX u , the terrain stability index of the u-th common disaster area is ZX u , the induced weight is ω1, the geological weight is ω2, the topographic weight is ω3, and the ecological weight is The building weight is ω5, and the building vulnerability of the u-th common disaster area is BR u ;

[0111] Calculate the comprehensive risk index of superstorm disaster areas:

[0112]

[0113] The comprehensive risk index of the u-th super rainstorm disaster area is The disaster area is marked according to the comprehensive risk index and the marked disaster area is output.

[0114] In this embodiment, the method for calculating the damage degree index of the disaster area includes:

[0115] Calculate the loss of life in the disaster area based on the data to be analyzed:

[0116]

[0117] The loss of life is LM s , the number of disaster areas is N a , the number of risk population is N p , the mortality rate of the risk population is η, and the influence of the bth major influencing factor on the mortality rate of the risk population is γ b , the weight coefficient of the bth main influencing factor is φ b , the influence of the bth secondary factor on the mortality rate of the risk population is ξ b , the weight coefficient of the bth secondary factor is The number of secondary factors is N b2 , the number of main factors is N b1 , the weight coefficient is β, and the correction coefficient is

[0118] Calculate the economic losses in the disaster area:

[0119]

[0120] The economic loss is D s The number of pixel disaster areas where properties are distributed is N1, the type of properties in the disaster area of ​​the vth pixel is N2, the number of different flooding depth levels is N3, and the loss rate of the vth type of property in the ath disaster area corresponding to the xth flooding depth is The property value of the vth type of property in the ath disaster area corresponding to the xth flooding depth is

[0121] Calculate the indirect economic losses in the disaster area:

[0122]

[0123] The indirect coefficient corresponding to the rth type of property is K a (s), the number of property categories is M1, and the rth property is A r , the indirect economic loss is

[0124] Calculate the ecological environment loss rate:

[0125]

[0126] The ecological environment loss rate is E, and the loss rate of the c-type ecosystem service function is Q c , the number of ecosystem service functions is N4, the number of ecological products is N5, and the service volume of the i-th product in the c-th ecosystem service function is H ck , the unit price of the i-th product in the c-th ecosystem service function is R ck ;

[0127] The damage index is calculated based on the loss of life, economic losses, indirect economic losses, and ecological environment loss rates:

[0128]

[0129] The total economic volume is KS, and the damage index is

[0130] In this embodiment, the method for performing real-time dynamic evaluation of the disaster area based on the damage degree index and the comprehensive risk index to obtain a loss index includes:

[0131] The damage degree index and comprehensive risk index are input into a real-time dynamic assessment model based on a Bayesian network, and the state-time curve is used to discretize the disaster state into three levels: extremely heavy rain, heavy rain, and torrential rain.

[0132] Taking time as a node, calculate the node's transition network conditional probability:

[0133]

[0134] The time interval is Δt, and the offset rate when node d changes from state w to state m is The state of the node at time t0 is w, and the state of the node at time t1 is j. Under the condition that the state is w at time t0, the probability of the node being in state j at time t1 is k(G t1 =j|G t0 =w), the damage index at time t0 is The comprehensive risk index at time t0 is The damage index of node d in state w is The damage index of node d in state m is The comprehensive risk index of node d in state w is The comprehensive risk index of node d in state m is

[0135] Calculate the loss index based on the conditional probability of the transfer network:

[0136]

[0137] The damage index of state j at time t is The comprehensive risk index in state j at time t is The loss index at time t is The damage index at time t in state w is The comprehensive risk index in state w at time t is

[0138] In this embodiment, a method for constructing a meta-learning-based rainstorm disaster loss assessment model based on the loss index includes:

[0139] The mean square error between actual rainstorm disaster loss assessment and predicted disaster loss is used as the assessment loss function, and the weighted addition of loss index and assessment loss function is used as the objective function of the rainstorm disaster loss assessment model.

[0140] Initialize the parameters of the rainstorm disaster loss assessment model and use the long short-term memory network as the predictor of the rainstorm disaster loss assessment model;

[0141] Randomly sample the data to be analyzed from the meta-training task, use the support set to calculate the inner gradient, and update the parameters of the rainstorm disaster loss assessment model. The expression is:

[0142]

[0143] The qth meta-training task is T q , the inner learning rate is α, and the initial parameters are Meta-training task T q The updated parameters of the inner layer are Meta-training task T q The loss function is about the initial parameters The gradient of Initial parameters The rainstorm disaster loss assessment model is

[0144] Compute the meta-loss using the support set:

[0145]

[0146] The loss of the The number of meta-training tasks is Using parameters on querysets Compute the meta-training task T q The loss is Update the outer parameters, the expression is:

[0147]

[0148] The outer layer learning rate is ψ, and the updated outer layer parameters are parameter The rainstorm disaster loss assessment model is Meta-loss with respect to parameters The gradient of

[0149] Output the rainstorm disaster loss assessment model after adding meta-learning.

[0150] The second aspect is the rainstorm disaster loss assessment system based on machine learning, which includes:

[0151] Data collection module: used to collect data from rainstorm disaster data sources in a preset area within a specified time period as data to be analyzed; the data sources to be analyzed include meteorological data, geological data, socioeconomic data, disaster data, social media data, and historical disaster data; the historical disaster data includes historical meteorological data, historical geological data, historical socioeconomic data, and historical disaster data;

[0152] Disaster area division module: used to divide the preset area into regions based on rainfall according to the data to be analyzed, and to divide the preset area with rainfall greater than a threshold into disaster areas, and vice versa as uncertain areas, and to calculate the destructive force of the uncertain area based on the hydrological mechanism and drainage system, and to add the uncertain area with rainfall greater than the threshold into the disaster area;

[0153] An analysis and real-time dynamic evaluation module is configured to perform a comprehensive risk analysis on the data to be analyzed in the disaster area to obtain a comprehensive risk index, calculate a damage degree index of the disaster area, and perform a real-time dynamic evaluation of the disaster area based on the damage degree index and the comprehensive risk index to obtain a loss index;

[0154] Modeling output module: used to construct a rainstorm disaster loss assessment model based on the loss index, input the data to be assessed into the rainstorm disaster loss assessment model, and output the assessment results.

[0155] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A rainstorm disaster loss assessment method based on machine learning, characterized by: The following steps are involved: Data from a rainstorm disaster data source in a preset area within a specified time period is used as the data to be analyzed; the data source to be analyzed includes meteorological data, geological data, topographic data, socioeconomic data, social media data, historical disaster data, ecological data, and architectural data; The method divides the preset area into regions based on the data to be analyzed based on rainfall, defines the preset area with a rainfall greater than a threshold as a disaster area, and defines the area with a rainfall greater than a threshold as an uncertain area, calculates the destructive force of the uncertain area based on the hydrological mechanism and the drainage system, and adds the uncertain area with a rainfall greater than the threshold to the disaster area; the method includes: Calculate the building's exposure, sensitivity, and adaptability based on building data, and calculate the destructive power through objective weighting: The building vulnerability of the i-th uncertain area is , the exposure of the i-th uncertainty area is , the sensitivity of the i-th uncertainty region is , the adaptability of the i-th uncertain region is , the exposure weight is , the sensitivity weight is , the adaptability is , the destructive force of the i-th uncertain region is , the rainfall in the i-th uncertain area is ; Performing a comprehensive risk analysis on the data to be analyzed in the disaster area to obtain a comprehensive risk index, calculating a damage degree index of the disaster area, and performing a real-time dynamic assessment of the disaster area based on the damage degree index and the comprehensive risk index to obtain a loss index; Constructing a rainstorm disaster loss assessment model based on meta-learning according to the loss index, inputting the data to be assessed into the rainstorm disaster loss assessment model, and outputting an assessment result; The method for performing comprehensive risk analysis on the data to be analyzed in the disaster area includes: The disaster-inducing factor index of common disaster areas is calculated through meteorological data, the geological vulnerability index of disaster areas is calculated through geological data, the terrain stability index is calculated through topographic and geomorphological data, and the building vulnerability is calculated through building data; Ecological sensitivity, ecological resilience, and ecological pressure are calculated through ecological data, and the ecological vulnerability index is calculated based on ecological sensitivity, ecological resilience, and ecological pressure: The ecological vulnerability index of the u-th common disaster area is , the ecological sensitivity is , the ecological pressure is , ecological resilience is , the sensitive weight is , the pressure weight is , the restoring force weight is ; Use disaster data and social media data to revise the disaster inducing factor index, geological vulnerability index, and terrain stability index; The comprehensive risk index of common rainstorm disaster areas is calculated based on the disaster inducing factor index, geological vulnerability index, terrain stability index and ecological vulnerability index: The comprehensive risk index of the u-th common disaster area is , the disaster inducing factor index of the u-th common disaster area is , the geological vulnerability index of the u-th common disaster area is , the terrain stability index of the u-th common disaster area is , the induced weight is , the geological weight is , the terrain weight is , the ecological weight is , the building weight is , the building vulnerability of the u-th general disaster area is ; Calculate the comprehensive risk index of superstorm disaster areas: The comprehensive risk index of the u-th super rainstorm disaster area is , risk marking of disaster areas is performed according to the comprehensive risk index, and the marked disaster areas are output.

2. The method for rainstorm disaster loss assessment based on machine learning according to claim 1 is characterized in that: The method for dividing the preset area according to the data to be analyzed based on rainfall includes: The data to be analyzed and the preset area are aligned in time and space, and the expert method is used to perform primary division of the preset area to obtain the initial area. The rainfall in the initial area is calculated based on the data to be analyzed. The initial area with a rainfall greater than the threshold is regarded as a disaster area, otherwise it is regarded as an uncertain area.

3. The method for rainstorm disaster loss assessment based on machine learning according to claim 1 is characterized in that: The method for the disaster area, comprising: The rainfall threshold is divided into ordinary rainfall threshold and super-threshold rainfall, and the disaster area is divided into ordinary disaster area and super-heavy rainfall disaster area; Disaster areas with rainfall amounts above the threshold for normal rainfall but less than the threshold for heavy rainfall are classified as normal disaster areas; disaster areas with rainfall amounts above the threshold for heavy rainfall are classified as super heavy rainfall disaster areas. Normal rainfall thresholds and above-threshold rainfall are dynamically adjusted according to the rainfall threshold drift due to climate warming.

4. The method for rainstorm disaster loss assessment based on machine learning according to claim 1 is characterized in that: The method for calculating the damage severity index of the disaster area includes: Calculate the loss of life in the disaster area based on the data to be analyzed: The loss of life is , the number of disaster areas is The number of people at risk is , the mortality rate of the risk population is , the influence degree of the bth main influencing factor on the mortality rate of the risk population is , the weight coefficient of the bth main influencing factor is , the influence of the bth secondary factor on the mortality rate of the risk population is , the weight coefficient of the bth secondary factor is , the number of secondary factors is , the number of main factors is , the weight coefficient is , the correction factor is ; Calculate the economic losses in the disaster area: The economic losses are , the number of pixels with property distribution in the disaster area is , the property type of the disaster area in the vth pixel is , the number of different flooding depth levels is The loss rate of the v-type property in the a-th disaster area corresponding to the x-th flooding depth is The property value of the vth type of property in the ath disaster area corresponding to the xth flooding depth is ; Calculate the indirect economic losses in the disaster area: The indirect coefficient corresponding to the rth type of property is , the number of property categories is , the rth type of property is , the indirect economic loss is ; Calculate the ecological environment loss rate: The ecological environment loss rate is The loss rate of ecosystem service function type c is , the number of ecosystem service functions is , the number of ecological products is , the service volume of the i-th product in the c-th ecosystem service function is , the unit price of the i-th product in the c-th ecosystem service function is ; The damage index is calculated based on the loss of life, economic losses, indirect economic losses, and ecological environment loss rates: The total economic output is The damage index is .

5. The method for rainstorm disaster loss assessment based on machine learning according to claim 1 is characterized in that: The method for performing real-time dynamic evaluation of the disaster area according to the damage degree index and the comprehensive risk index to obtain a loss index includes: The damage degree index and comprehensive risk index are input into a real-time dynamic assessment model based on a Bayesian network, and the state-time curve is used to discretize the disaster state into three levels: extremely heavy rain, heavy rain, and torrential rain. Taking time as a node, calculate the node's transition network conditional probability: The time interval is , the offset rate of node d when it changes from w state to m state is , The state of the node at this moment is w, The state of the node at time j is Under the condition of w state at the moment The probability of being in state j at time is , The damage index at the moment is , The comprehensive risk index at the moment is , the damage index of node d in state w is , the damage index of node d in state m is , the comprehensive risk index of node d in state w is , the comprehensive risk index of node d in state m is ; Calculate the loss index based on the conditional probability of the transfer network: The damage index of state j at time t is , the comprehensive risk index in state j at time t is , the loss index at time t is , the damage index at time t in state w is , the comprehensive risk index in state w at time t is .

6. The method for rainstorm disaster loss assessment based on machine learning according to claim 1 is characterized in that: A method for constructing a meta-learning-based rainstorm disaster loss assessment model based on the loss index includes: The mean square error between actual rainstorm disaster loss assessment and predicted disaster loss is used as the assessment loss function, and the weighted addition of loss index and assessment loss function is used as the objective function of the rainstorm disaster loss assessment model. Initialize the parameters of the rainstorm disaster loss assessment model and use the long short-term memory network as the predictor of the rainstorm disaster loss assessment model; Randomly sample the data to be analyzed from the meta-training task, use the support set to calculate the inner gradient, and update the parameters of the rainstorm disaster loss assessment model. The expression is: The qth meta-training task is , the inner learning rate is , the initial parameters are , meta-training task The updated parameters of the inner layer are , meta-training task The loss function is about the initial parameters The gradient of , initial parameters The rainstorm disaster loss assessment model is ; Compute the meta-loss using the support set: The loss of the , the number of meta-training tasks is , using parameters on querysets Computational meta-training tasks The loss is ; Update the outer parameters, the expression is: The outer learning rate is , the updated outer parameters are ,parameter The rainstorm disaster loss assessment model is , the meta-loss is about the parameters The gradient of ; Output the rainstorm disaster loss assessment model after adding meta-learning.

7. A rainstorm disaster loss assessment system based on machine learning, used to implement the method according to any one of claims 1 to 6, characterized in that: include: Data collection module: used to collect data from rainstorm disaster data sources in a preset area within a specified time period as data to be analyzed; the data sources to be analyzed include meteorological data, geological data, socioeconomic data, disaster data, social media data, and historical disaster data; the historical disaster data includes historical meteorological data, historical geological data, historical socioeconomic data, and historical disaster data; Disaster area division module: used to divide the preset area into regions based on rainfall according to the data to be analyzed, and to divide the preset area with rainfall greater than a threshold into disaster areas, and vice versa as uncertain areas, and to calculate the destructive force of the uncertain area based on the hydrological mechanism and drainage system, and to add the uncertain area with rainfall greater than the threshold into the disaster area; An analysis and real-time dynamic evaluation module is configured to perform a comprehensive risk analysis on the data to be analyzed in the disaster area to obtain a comprehensive risk index, calculate a damage degree index of the disaster area, and perform a real-time dynamic evaluation of the disaster area based on the damage degree index and the comprehensive risk index to obtain a loss index; Modeling output module: used to construct a rainstorm disaster loss assessment model based on the loss index, input the data to be assessed into the rainstorm disaster loss assessment model, and output the assessment results.

Images