Bayesian network-based rainstorm-mountain torrent disaster chain trigger threshold identification method

Abstract

The invention discloses a Bayesian network-based rainstorm-mountain torrent disaster chain trigger threshold identification method, which comprises the following steps: according to a preset fuzzy character string matching algorithm and a preset text similarity algorithm, carrying out cleaning processing on extracted historical disaster situation texts in sequence to obtain processed disaster situation texts; performing feature extraction on the processed disaster text to obtain key disaster features; constructing a target Bayesian network topology of the rainstorm-mountain torrent disaster chain; discretizing each node variable in the target Bayesian network topology to obtain a state value of the node variable; the prior probability of the node variables is calculated, and the conditional probability of the causal relationship between the nodes is calculated according to the prior probability and a preset application expectation maximization algorithm; calculating a target posterior probability that the first disaster triggers the derived disaster; and according to a preset posterior probability critical value and the target posterior probability, inverting the intensity of the first disaster, and determining an intensity threshold value of the derived disaster triggered by the first disaster.

Description

Technical Field

[0001] The present invention relates to the technical field of natural disaster management, and in particular to a method for identifying the triggering threshold of a rainstorm-mountain flood disaster chain based on a Bayesian network. Background Art

[0002] Under the background of current global climate change, extreme rainstorm events occur frequently and widely, and mountain flood disasters also show an increasing trend in occurrence frequency and an expanding influence range. In addition, mountain flood disasters often breed and evolve in environments with heavy rainfall, large terrain slopes, and loose soil quality. During the disaster-causing process, derivative disasters such as landslides and debris flows will be triggered. Therefore, under special weather and topographical and geomorphic conditions, there is a time-series causal and chain-triggering relationship among rainstorms, mountain floods, landslides, and debris flows, which is likely to induce a rainstorm-mountain flood disaster chain and drive the adverse situations of "multiple disasters in one place" and "amplification of disaster conditions in a chain".

[0003] Currently, although there are many studies on natural disaster databases such as rainstorms, mountain floods, landslides, and debris flows at home and abroad, in terms of the extensiveness of disaster situation data collection and the integrity of disaster feature extraction, the existing technologies have the following deficiencies: First, the previous disaster situation data came from government department bulletins, with a single information source, serious fragmentation and lag, and it was difficult to obtain a large amount of timely rainstorm, mountain flood, landslide, and debris flow disaster data, resulting in untimely updates in the disaster situation database and being unable to reflect the latest dynamics of disasters in a timely manner; Second, the analysis methods for massive disaster situation data on social media such as Weibo are not yet perfect, and it is difficult to quickly extract key disaster features such as the time of disaster occurrence, disaster intensity, influence range, casualties, and property losses from them, thus reducing the accuracy and integrity of the disaster situation database. Third, the rainstorm-mountain flood disaster chain generally breaks out under the impact of high-intensity rainstorms. When the rainstorm exceeds what intensity threshold, it will trigger derivative disasters such as mountain floods, landslides, and debris flows? The triggering mechanism problem hidden behind the rainstorm-mountain flood disaster chain phenomenon has not yet been revealed.

[0004] Therefore, there is an urgent need for a suitable method to create a historical disaster situation database of the rainstorm-mountain flood disaster chain and identify the triggering threshold of the rainstorm-mountain flood disaster chain, providing important technical support for the early warning and collaborative prevention and control of complex chain disasters. Summary of the Invention

[0005] In order to solve the above problems existing in the prior art, the present invention provides a method for identifying the triggering threshold of a rainstorm-mountain flood disaster chain based on a Bayesian network, so as to solve the technical problem that in the prior art, it has not been answered what intensity threshold the rainfall exceeds to trigger derivative disasters such as mountain floods, landslides, and debris flows, that is, the research on the triggering mechanism of the rainstorm-mountain flood disaster chain is still relatively weak.

[0006] In order to achieve the above object, the technical solution of the embodiment of the present invention is:

[0007] In a first aspect, the present invention provides a method for identifying the triggering threshold of a rainstorm-mountain flood disaster chain based on a Bayesian network, the method comprising:

[0008] Extracting historical disaster situation texts related to the rainstorm-mountain flood disaster chain from multiple data sources;

[0009] According to a preset fuzzy string matching algorithm and a preset text similarity algorithm, sequentially cleaning the historical disaster situation texts to obtain processed disaster situation texts;

[0010] Performing feature extraction on the processed disaster situation texts to obtain key disaster features;

[0011] Constructing the target Bayesian network topology of the rainstorm-mountain flood disaster chain according to the processed disaster situation texts and the key disaster features;

[0012] Performing discretization processing on each node variable in the target Bayesian network topology to obtain the state values of the node variables; the state values represent the occurrence and non-occurrence states of disaster events;

[0013] Calculating the prior probability of the node variable, and calculating the conditional probability of the causal relationship between nodes according to the prior probability, the state values, and a preset application expectation maximization algorithm;

[0014] Calculating the target posterior probability of the triggering of a derivative disaster by a prior disaster;

[0015] According to a preset posterior probability critical value and the target posterior probability, inversely inferring the intensity of the prior disaster to determine the intensity threshold for the prior disaster to trigger the derivative disaster.

[0016] In a second aspect, the present invention provides a device for identifying the triggering threshold of a rainstorm-mountain flood disaster chain based on a Bayesian network, the device comprising:

[0017] An extraction module, configured to extract historical disaster situation texts related to the rainstorm-mountain flood disaster chain from multiple data sources;

[0018] A processing module, configured to sequentially clean the historical disaster situation texts according to a preset fuzzy string matching algorithm and a preset text similarity algorithm to obtain processed disaster situation texts;

[0019] The extraction module is further configured to perform feature extraction on the processed disaster situation texts to obtain key disaster features;

[0020] A construction module, configured to construct the target Bayesian network topology of the rainstorm-mountain flood disaster chain according to the processed disaster situation texts and the key disaster features;

[0021] A discretization module for discretizing each node variable in the target Bayesian network topology to obtain state values of the node variables; the state values represent the occurrence and non-occurrence states of disaster events.

[0022] A calculation module for calculating the prior probability of the node variable and calculating the conditional probability of the causal relationship between nodes according to the prior probability, the state values, and a preset application of the expectation maximization algorithm.

[0023] The calculation module is further configured to calculate the target posterior probability of the triggering of a derivative disaster by a prior disaster.

[0024] An inversion module for inverting the intensity of the prior disaster according to a preset posterior probability threshold and the target posterior probability, and determining the intensity threshold for the triggering of the derivative disaster by the prior disaster.

[0025] In some embodiments, the processing module is further configured to calculate the edit distance between two historical disaster situation texts according to the preset fuzzy string matching algorithm; determine the text similarity between the two historical disaster situation texts according to the edit distance and the preset text similarity algorithm; the text similarity is determined by the following formula: ; where The distance is the minimum number of single-character edit operations required to convert one string to another string. And Are the historical disaster situation texts 1 and 2 to be compared respectively. Represents the length function of the string; delete the historical disaster situation texts with text similarity exceeding the preset similarity threshold in the historical disaster situation texts to obtain the processed disaster situation texts; extract features from the processed disaster situation texts to obtain the key disaster features.

[0026] In some embodiments, the construction module is further configured to integrate the information of the processed disaster situation texts and the key disaster features according to the disaster type and the disaster occurrence time, establish the rainstorm-mountain flood disaster chain database; determine the influencing factors of the rainstorm-mountain flood disaster chain according to the rainstorm-mountain flood disaster chain database; construct the target Bayesian network topology with the influencing factors and the key disaster features as node variables and the causal relationship between the node variables as directed edges.

[0027] In a third aspect, an embodiment of the present invention provides an electronic device, including: a memory for storing executable instructions; a processor for implementing the above-mentioned method for identifying the triggering threshold of a rainstorm-mountain flood disaster chain based on a Bayesian network when executing the executable instructions stored in the memory.

[0028] Fourthly, an embodiment of the present invention provides a computer-readable storage medium storing executable instructions, which, when causing a processor to execute the executable instructions, implement the above-mentioned method for identifying the triggering threshold of the rainstorm-mountain flood disaster chain based on a Bayesian network.

[0029] The method for identifying the triggering threshold of the rainstorm-mountain flood disaster chain based on a Bayesian network provided by the present invention includes: First, according to a preset fuzzy string matching algorithm and a preset text similarity algorithm, the extracted historical disaster situation texts are sequentially cleaned to obtain the processed disaster situation texts; feature extraction is performed on the processed disaster situation texts to obtain key disaster features; the target Bayesian network topology of the rainstorm-mountain flood disaster chain is constructed; each node variable in the target Bayesian network topology is discretized to obtain the state values of the node variables; the prior probability of the node variables is calculated, and according to the prior probability and a preset application of the expectation maximization algorithm, the conditional probability of the causal relationship between nodes is calculated; the target posterior probability of the triggering of a derivative disaster by a prior disaster is calculated; according to a preset posterior probability critical value and the target posterior probability, the intensity of the prior disaster is inverted to determine the intensity threshold for the prior disaster to trigger a derivative disaster. In this way, the present invention systematically clarifies the inducing causes of the rainstorm-mountain flood disaster chain, that is, under the impact of a certain high-intensity rainstorm, derivative disasters such as mountain floods, landslides, and debris flows will be triggered; in addition, it also clarifies the triggering mechanism hidden behind the rainstorm-mountain flood disaster chain phenomenon; and promotes the transformation of flood control from single-disaster response to comprehensive disaster reduction, improves the early warning ability of mountain flood disasters, improves the overall efficiency of disaster prevention and mitigation, reduces the losses caused by mountain flood disasters, and can also provide strong data support for policymakers and managers, providing important scientific support for the sustainable development of social economy and environmental ecology. Description of the Drawings

[0030] Figure 1 is a schematic structural diagram of a system for identifying the triggering threshold of a rainstorm-mountain flood disaster chain based on a Bayesian network provided by an embodiment of the present invention;

[0031] Figure 2 is a schematic flow diagram of a method for identifying the triggering threshold of a rainstorm-mountain flood disaster chain based on a Bayesian network provided by an embodiment of the present invention;

[0032] Figure 3 is a schematic flow diagram of a method for identifying the triggering threshold of a rainstorm-mountain flood disaster chain provided by an embodiment of the present invention;

[0033] Figure 4 is a schematic composition structure diagram of a device for identifying the triggering threshold of a rainstorm-mountain flood disaster chain based on a Bayesian network provided by an embodiment of the present invention;

[0034] Figure 5It is a schematic diagram of the composition structure of the rainstorm - mountain flood disaster chain trigger threshold identification device provided by the embodiments of the present invention. Detailed implementation manners

[0035] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention.

[0036] In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. Unless otherwise defined, all technical and scientific terms used in the embodiments of the present invention have the same meaning as commonly understood by those skilled in the technical field to which the embodiments of the present invention belong. The terms used in the embodiments of the present invention are only for the purpose of describing the embodiments of the present invention and are not intended to limit the present invention.

[0037] The following describes the exemplary application of the rainstorm - mountain flood disaster chain trigger threshold identification device based on the Bayesian network according to the embodiments of the present invention. The rainstorm - mountain flood disaster chain trigger threshold identification device provided by the embodiments of the present invention can be implemented as a terminal or a server. In one implementation, the rainstorm - mountain flood disaster chain trigger threshold identification device provided by the embodiments of the present invention can be implemented as various types of terminals such as laptops, tablets, desktop computers, mobile devices, etc.; in another implementation, the rainstorm - mountain flood disaster chain trigger threshold identification device provided by the embodiments of the present invention can also be implemented as a server. Among them, the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery network (CDN, Content Delivery Network), and big data and artificial intelligence platforms. The terminal and the server can be directly or indirectly connected through wired or wireless communication methods, which are not limited in the embodiments of the present invention. Below, the exemplary application when the rainstorm - mountain flood disaster chain trigger threshold identification device based on the Bayesian network is a server will be described.

[0038] See Figure 1 , Figure 11 is a schematic diagram of the structure of the Bayesian network-based rainstorm-flash flood disaster chain trigger threshold identification system 10 provided in an embodiment of the present invention. In order to realize the identification of the rainstorm-flash flood disaster chain trigger threshold, the embodiment of the present invention can provide a rainstorm-flash flood disaster chain trigger threshold identification platform based on the Bayesian network, and the rainstorm-flash flood disaster chain trigger threshold identification platform based on the Bayesian network can be implemented as a rainstorm-flash flood disaster chain trigger threshold identification application based on the Bayesian network. The Bayesian network-based rainstorm-flash flood disaster chain trigger threshold identification system 10 provided in an embodiment of the present invention includes a terminal 110, a network 120 and a server 130, wherein the server 130 is a server of the Bayesian network-based rainstorm-flash flood disaster chain trigger threshold identification application. The server 130 can constitute the Bayesian network-based rainstorm-flash flood disaster chain trigger threshold identification device of the embodiment of the present invention. The terminal 110 is connected to the server 130 through the network 120, and the network 120 can be a wide area network or a local area network, or a combination of the two.

[0039] In some embodiments, please refer to Figure 1 When the trigger threshold of the rainstorm-flash flood disaster chain is identified, the terminal 110 sends the rainstorm-flash flood disaster chain trigger threshold identification task to the server 130 through the network 120. In response to the rainstorm-flash flood disaster chain trigger threshold identification task sent by the terminal 110, the server 130 extracts historical disaster texts related to the rainstorm-flash flood disaster chain from multiple data sources; according to the preset fuzzy string matching algorithm and the preset text similarity algorithm, the historical disaster texts are cleaned in turn to obtain the processed disaster texts; the processed disaster texts are feature extracted to obtain key disaster features; according to the processing The target Bayesian network topology of the rainstorm-flash flood disaster chain is constructed by using the processed disaster text and key disaster features; each node variable in the target Bayesian network topology is discretized to obtain the state value of the node variable; the prior probability of the node variable is calculated, and the conditional probability of the causal relationship between the nodes is calculated based on the prior probability and the preset application expectation maximization algorithm; the target posterior probability of the first disaster triggering the derivative disaster is calculated; based on the preset posterior probability critical value and the target posterior probability, the intensity of the first disaster is inverted to determine the intensity threshold of the first disaster triggering the derivative disaster, that is, the derivative disaster triggering intensity threshold is obtained. After obtaining the derivative disaster triggering intensity threshold, the server 130 sends the derivative disaster triggering intensity threshold to the terminal 110 through the network 120.

[0040] The embodiment of the present invention provides a method for identifying the trigger threshold of a rainstorm-flash flood disaster chain based on a Bayesian network. Figure 2 , Figure 2 is a flow chart of a method for identifying a trigger threshold of a rainstorm-flash flood disaster chain based on a Bayesian network provided by an embodiment of the present invention, which is combined with Figure 2The steps shown are explained.

[0041] Step S210, extracting historical disaster texts related to the rainstorm-flash flood disaster chain from multiple data sources.

[0042] In some embodiments, a rainstorm-flash flood disaster chain refers to a series of disasters caused by rainstorms, in which rainstorms serve as the initial disaster, triggering derivative disasters such as flash floods, forming a disaster sequence with a causal relationship.

[0043] In some embodiments, historical disaster text refers to textual materials that record disaster situations related to the rainstorm-flash flood disaster chain that occurred in the past, which may come from multiple data sources such as news reports, government disaster reports, scientific research reports, etc. It may include information such as the time, location, degree of damage, and manifestation of the disaster.

[0044] In the present invention, text contents directly or indirectly related to the rainstorm-flash flood disaster chain are screened out from various places that may contain relevant information, such as meteorological department records, disaster report files, news reports, field investigation materials and other data sources, to provide basic data for subsequent analysis.

[0045] Step S220, according to a preset fuzzy string matching algorithm and a preset text similarity algorithm, the historical disaster text is cleaned in sequence to obtain a processed disaster text.

[0046] In some embodiments, the preset fuzzy string matching algorithm refers to a pre-set algorithm for finding content similar to a specific pattern or string in a text, which allows a certain degree of fuzziness and can handle situations where the match is not completely accurate, and is used to screen and locate relevant information in historical disaster texts. For example, edit distance algorithm, Jaccard similarity algorithm, regular expression matching algorithm, etc. For example, some text descriptions are "heavy rains cause urban flooding", while others may be expressed as "due to heavy rains, urban flooding occurs", and the preset fuzzy string matching algorithm can recognize such similar semantic expressions. Common fuzzy string matching algorithms include algorithms based on character editing operations (such as inserting, deleting, and replacing characters), the core of which is to measure the similarity between two strings by calculating the minimum number of editing operations required to convert one string into another.

[0047] In some embodiments, the preset text similarity algorithm refers to an algorithm that evaluates the similarity degree of two texts at the content and semantic levels through a quantitative method. It can analyze based on features such as the vocabulary usage, grammatical structure, and semantic understanding of the text, and calculate a numerical value to represent the similarity. The closer the numerical value is to 1, the higher the text similarity; the closer it is to 0, the lower the similarity, thereby judging the degree of closeness of the relationship between texts. In the present invention, the preset fuzzy string matching algorithm can be used to sort out the collected historical disaster situation texts, identify and correct the incorrect expressions in the texts, such as standardizing different writings of "flash flood" into the standard expression; removing redundant content that appears repeatedly to avoid data interference; unifying the text formats, such as unifying the date format, number format, etc. Through these operations, the accuracy and standardization of the text data are improved, providing a reliable data basis for subsequent analysis. Then, using the preset text similarity algorithm, the cleaned text is deeply analyzed. Key feature information closely related to the rainstorm-mountain flood disaster chain is extracted, such as the precise time of the disaster occurrence, the specific geographical location, the rainstorm intensity level (such as light rain, moderate rain, heavy rain, etc.), the scale of the mountain flood (such as the size of the flood flow, the inundation range), the affected objects and the degree of damage (such as the number of damaged houses, the affected area of crops, the casualty situation of people), etc. These features are accurately extracted from the original text and sorted out to form the processed disaster situation text for further mining and analysis of key information in the future.

[0048] Step S230, extract features from the processed disaster situation text to obtain key disaster features.

[0049] In some embodiments, the key disaster features refer to the features extracted from the historical disaster situation text that can represent the key information of the rainstorm-mountain flood disaster chain.

[0050] Step S240, according to the processed disaster situation text and the key disaster features, construct the target Bayesian network topology of the rainstorm-mountain flood disaster chain.

[0051] In some embodiments, the target Bayesian network topology refers to a graphical model based on probability reasoning calculated through the above steps, used to represent the causal relationship between variables, and describes the connection method and dependency relationship between these variables.

[0052] In the present invention, based on the processed disaster situation text and the key disaster features, the node variables in the Bayesian network and their connection relationships are determined, and a graphical structure that can represent the causal relationship between various factors in the rainstorm-mountain flood disaster chain is constructed, that is, the target Bayesian network topology. For example, taking the rainstorm intensity as one node and the occurrence situation of the mountain flood as another node, and determining their causal connection by analyzing historical data.

[0053] Step S250: Discretize each node variable in the target Bayesian network topology to obtain the state values of the node variables; the state values represent the occurrence and non-occurrence states of disaster events.

[0054] In the present invention, node variables represent various features or variables related to the rainstorm - mountain flood disaster chain. Here, node variables can at least include the influencing factors and key disaster situation features of the rainstorm - mountain flood disaster chain. Here, the influencing factors mainly include: total precipitation, annual average precipitation, rainfall intensity, precipitation duration (the time of a single precipitation), annual average rainfall, soil moisture, DEM elevation, vegetation coverage rate, slope, landslides, debris flows, casualty situations, number of collapsed houses, affected area of crops, etc.

[0055] In the present invention, continuous node variables in the target Bayesian network topology, such as the specific value of rainstorm intensity, specific data of mountain flood discharge, etc., are divided into discrete states according to certain rules. For example, the rainstorm intensity is divided into several levels such as "light rain", "moderate rain", "heavy rain", "rainstorm", etc. Each level is a state value of the node variable, which is convenient for probability calculation and model analysis.

[0056] Step S260: Calculate the prior probability of the node variable, and calculate the conditional probability of the causal relationship between nodes according to the prior probability, the state value, and the preset expectation maximization algorithm.

[0057] In the present invention, according to the frequency and statistical rules of each node variable appearing in historical disaster situation texts, estimate the probability of each node variable being in different state values without other conditional restrictions. For example, according to past records, calculate the probability that the rainstorm intensity in a certain area is at the "rainstorm" level. Using the preset expectation maximization algorithm, combined with the prior probability and the causal relationship between nodes, calculate the probability that another node variable with a causal relationship is in a different state when a node variable is in a certain state. For example, calculate the probability of a mountain flood occurring when the rainstorm intensity is at the "rainstorm" level.

[0058] Step S270: Calculate the target posterior probability of the triggering of derivative disasters by precursor disasters.

[0059] In the present invention, after determining the prior probability of node variables and the conditional probability of the causal relationship between nodes, according to Bayes' formula and specific disaster situation information, calculate the probability that a precursor disaster (such as a rainstorm) triggers a derivative disaster (such as a mountain flood) under certain known conditions, that is, the target posterior probability. This probability can more accurately reflect the likelihood of a rainstorm triggering a mountain flood in actual situations.

[0060] Step S280: Invert the intensity of the antecedent disaster based on the preset posterior probability threshold and the target posterior probability, and determine the intensity threshold at which the antecedent disaster triggers the derivative disaster.

[0061] In the present invention, the calculated target posterior probability is compared with the preset posterior probability threshold. If the target posterior probability exceeds the threshold, it indicates that in the current situation, the possibility of the antecedent disaster triggering the derivative disaster is relatively high. Through analysis and calculation, the intensity that the antecedent disaster (such as heavy rain) needs to reach at this time is inversely deduced, and this intensity is the intensity threshold at which the antecedent disaster triggers the derivative disaster. For example, when the target posterior probability indicates a high possibility of mountain flood occurrence, determine the corresponding heavy rain intensity at this time as the intensity threshold for triggering the mountain flood, which has important guiding significance for disaster warning and prevention.

[0062] The method for identifying the triggering threshold of the rainstorm-mountain flood disaster chain based on the Bayesian network provided by the present invention first, according to the preset fuzzy string matching algorithm and the preset text similarity algorithm, sequentially perform cleaning processing on the extracted historical disaster situation texts to obtain the processed disaster situation texts; extract features from the processed disaster situation texts to obtain key disaster features; construct the target Bayesian network topology of the rainstorm-mountain flood disaster chain; perform discretization processing on each node variable in the target Bayesian network topology to obtain the state values of the node variables; calculate the prior probability of the node variables, and calculate the conditional probability of the causal relationship between nodes according to the prior probability and the preset application of the expectation maximization algorithm; calculate the target posterior probability of the antecedent disaster triggering the derivative disaster; invert the intensity of the antecedent disaster based on the preset posterior probability threshold and the target posterior probability, and determine the intensity threshold at which the antecedent disaster triggers the derivative disaster. In this way, the present invention systematically clarifies the inducing causes of the rainstorm-mountain flood disaster chain, that is, under the impact of a certain high-intensity rainstorm, derivative disasters such as mountain floods, landslides, and debris flows will be triggered; in addition, it also clarifies the triggering mechanism hidden behind the rainstorm-mountain flood disaster chain phenomenon; and promotes the transformation of flood control from single-disaster response to comprehensive disaster reduction, improves the early warning ability of mountain flood disasters, improves the overall efficiency of disaster prevention and reduction, reduces the losses caused by mountain flood disasters, and can also provide strong data support for policymakers and managers, and provide important scientific support for the sustainable development of social economy and environmental ecology.

[0063] In some embodiments, the above step S220 further includes the following steps S221 to S224:

[0064] Step S221: Calculate the edit distance between two historical disaster situation texts according to the preset fuzzy string matching algorithm.

[0065] In some embodiments, the edit distance refers to the minimum number of single-character edit operations (insertion, deletion, replacement) required to convert one string into another between two strings. In the processing of historical disaster situation texts, the edit distance is used to quantify the degree of difference between two texts. The smaller the edit distance, the higher the similarity of the two historical disaster situation texts at the character level.

[0066] Step S222: Determine the text similarity between the pairs of historical disaster situation texts according to the edit distance and the preset text similarity algorithm.

[0067] Here, the text similarity is determined by the following formula:

[0068] ; where The distance is the minimum number of single-character edit operations required to convert one string into another; and are the historical disaster situation texts 1 and 2 to be compared respectively; represents the length function of the string.

[0069] Step S223: Delete the historical disaster situation texts in the historical disaster situation texts whose text similarity exceeds the preset similarity threshold to obtain the processed disaster situation texts.

[0070] In some embodiments, the preset similarity threshold is a critical value artificially preset for determining whether the text similarity is high enough. When processing historical disaster situation texts, after obtaining the similarity values of pairs of texts through the preset text similarity algorithm, a standard is needed to decide which texts are too similar and need to be deleted. For example, if the preset similarity threshold is set to 0.8, then when the calculated result of the text similarity of two historical disaster situation texts is greater than 0.8, it is considered that the text similarity of these two texts exceeds the preset threshold, and one of the texts may be deleted to avoid data redundancy. The setting of this threshold needs to be adjusted according to the actual application scenario and the requirement for data deduplication.

[0071] Step S224: Extract features from the processed disaster situation texts to obtain the key disaster features.

[0072] In some embodiments, the above step S240 further includes the following steps S241 to S243:

[0073] Step S241: Integrate the information of the processed disaster situation texts and the key disaster features according to the disaster type and the disaster occurrence time to establish the rainstorm-mountain flood disaster chain database.

[0074] Step S242: Determine the influencing factors of the rainstorm-mountain flood disaster chain according to the rainstorm-mountain flood disaster chain database.

[0075] Step S243: Construct the topology of the target Bayesian network with the influencing factors and the key disaster characteristics as node variables and the causal relationships between the node variables as directed edges.

[0076] Next, an exemplary application of the embodiments of the present invention in a practical application scenario will be described.

[0077] This embodiment proposes a method for identifying the triggering threshold of a rainstorm-mountain flood disaster chain based on a Bayesian network, including the following:

[0078] 1. Steps for creating a rainstorm-mountain flood disaster chain database based on Internet big data mining and generative AI, and the specific method is as follows:

[0079] 1.1. Use government department bulletins (National Disaster Reduction Network, China Disaster Prevention Association, Water and Drought Defense Bulletin, emergency management departments of each province, etc.), domestic and foreign disaster databases (Global Disaster Data Platform, EM-DAT Emergency Disaster Database, Global Compound Disaster Dataset, etc.), news information (China News Network, Sohu.com, Toutiao.com, etc.), and social media (Weibo, etc.) as data sources to conduct investigations on historical rainstorm, mountain flood, landslide, and debris flow disaster situations.

[0080] 1.2. Set the keyword format as "district / county-level city name + (rainstorm, debris flow, flood, landslide, etc.)", such as "Rainstorm in Fangshan District", "Mountain flood in Fangshan District", "Landslide in Fangshan District", "Debris flow in Fangshan District", etc. Use the Octopus data collector to classify and collect historical disaster situation texts of rainstorms, mountain floods, landslides, and debris flows from 2010 to 2024 from the four types of data sources.

[0081] 1.3. For the problems of information redundancy and lack of data description in the collected original disaster situation texts, perform data cleaning on them. Mainly use the fuzzy string matching algorithm based on the edit distance (Levenshtein Distance) to delete historical disaster situation texts with a similarity exceeding 80% to reduce the redundancy of disaster situation information. Among them, the similarity calculation formula is as follows:

[0082] ;

[0083] In the formula, Distance is an index to measure the difference degree between two strings, that is, the minimum number of single-character editing operations (insertion, deletion, replacement) required to convert one string into another string; And are the original disaster situation texts 1 and text 2 to be compared respectively; Represents the length function of a string.

[0084] 1.4. Combine the information extraction technology in natural language processing (NLP) with generative AI, call Spark Lite cognitive model, process and compile the cleaned historical disaster text, and extract the key disaster characteristics such as the time of occurrence, disaster intensity, impact range, casualties, and property losses of rainstorms, mountain torrents, landslides, and mudslides one by one. In the Spark cognitive model, the question template is defined as "question = How long did it take to occur? How many people were affected? How many deaths and missing persons were there? How many people were relocated? How many crops were affected? How many houses collapsed? How much was the direct economic loss (amount)? How much was the indirect economic loss (amount)?" The response format is "Time of occurrence: not mentioned; number of people affected: not mentioned; number of deaths and missing persons: not mentioned; number of relocated people: not mentioned; area of ​​crops affected: not mentioned; number of collapsed houses: not mentioned; direct economic loss: not mentioned; indirect economic loss: not mentioned".

[0085] 1.5. Integrate the extracted historical disaster information and its key disaster characteristics according to the disaster type and occurrence time, establish a structured database (spreadsheet) and spatial database (ArcGIS), sort the disaster data in the same area according to the order of "occurrence time", and extract all disaster chains. Count the number of disaster chains in the entire basin and identify the hot spots of heavy rain-flash flood disaster chains.

[0086] In this embodiment, in the big data era, records of rainstorms, mountain torrents, landslides and mud-rock flow events and their disaster impacts are not limited to historical documents, but can also be stored, disseminated and more easily obtained through the Internet platform. However, the historical rainstorm, mountain torrents, landslides and mud-rock flow disaster information on the Internet platform is scattered and complicated. Internet big data mining technology is used to collect these disaster texts, and natural language processing technology - calling the generative AI (Spark Cognitive Big Model) interface method to extract key disaster features, so as to efficiently create a historical disaster database of the rainstorm-mountain torrent disaster chain.

[0087] 2. The trigger threshold identification steps of the rainstorm-flash flood disaster chain based on Bayesian network are as follows:

[0088] 2.1. Combine expert opinions with literature research, and use the rainstorm-mountain flood disaster chain database created above to determine the influencing factors of the rainstorm-mountain flood disaster chain, mainly including: total precipitation, average annual precipitation, rainfall intensity, precipitation duration (the time of a single precipitation), average annual rainfall, soil moisture, DEM elevation, vegetation coverage rate, slope, landslide, debris flow, casualty situation, number of collapsed houses, affected area of crops, etc. Taking the influencing factors of the rainstorm-mountain flood disaster chain, key disaster situation characteristics, etc. as nodes, and the causal relationship between nodes as directed edges, construct the Bayesian network topology of the rainstorm-mountain flood disaster chain to lay a foundation for the probability inference of the rainstorm-mountain flood disaster chain.

[0089] 2.2. Considering that if the node variables are continuous, the parameter estimation of the conditional probability distribution of the node variables will become very complex. To reduce the complexity of parameter estimation, based on the typical disaster events in the rainstorm-mountain flood disaster chain database, discretize each node variable in the Bayesian network to obtain the state values of the node variables (two states of the occurrence and non-occurrence of the disaster event). Then, calculate the prior probability of the node variables in the way of replacing probability with frequency; and apply the expectation maximization (EM) algorithm for parameter learning to obtain the conditional probability indicating the causal relationship between nodes. The calculation formula is as follows:

[0090] ;

[0091] In the formula, is the child node in the Bayesian network topology, representing different types of derivative disasters; represents the different state values of node , which are discretized into states in total; represents the set of all parent nodes that affect the child node (there are parent nodes in total), where represents the th parent node; represents the conditional probability that the child node is in the state under the influence of the parent node .

[0092] 2.3. Take a certain derivative disaster that has occurred (such as mountain flood, landslide, debris flow) as a known variable (evidence variable), use the Bayesian inference method to calculate the posterior probability of its parent nodes, and characterize the probability that this derivative disaster (child node variable) is triggered by the prior disaster (parent node variable). And use the Berier standard test method to evaluate the reliability of the Bayesian network probability inference. Among them, the calculation formula of the posterior probability is as follows:

[0093] ;

[0094] In the formula, represents the prior probability of the parent node ; represents the state of the child node triggered by the parent node , that is, the posterior probability. The Berier standard test formula is as follows:

[0095] ;

[0096] In the formula, represents the average deviation value of the Bayesian network, The value range of is . When is smaller, the deviation of the constructed Bayesian network is smaller, and the reliability of probability reasoning is better. Indicates that the reliability meets the requirements; otherwise, it does not meet the requirements.

[0097] 2.4. In the constructed Bayesian network model, due to the different conditional probabilities of derivative disasters under different intensities of precursor disasters. Therefore, a sufficiently large posterior probability critical value is set, and the intensity of the precursor disaster is inversely deduced to determine the intensity threshold of the precursor disaster triggering the derivative disaster. Referring to the probability classification standard of climate extreme events of IPCC (see Table 1), when the probability exceeds 66%, 90% and 99%, it corresponds to possible, very likely and almost certain to occur respectively. Therefore, the critical value of the posterior probability is set to 90%, and the rainfall intensities that are very likely to cause flash floods, landslides and debris flows are calculated respectively by inverse deduction, and used as the intensity thresholds for rainstorms to trigger flash floods, landslides and debris flows.

[0098] Table 1 Uncertainty levels of the occurrence probability of extreme climate events

[0099]

[0100] This embodiment provides a method for identifying the triggering threshold of a rainstorm-flash flood disaster chain, as Figure 3 shown, the specific steps are as follows:

[0101] Step 1: First, use government bulletins, disaster databases, news information, and social media as four major data sources to conduct historical rainstorm, flash flood, landslide, and mud-rock flow disaster investigations. Then, use Internet big data mining technology to classify and collect disaster texts such as rainstorm, flash flood, landslide, and mud-rock flow. Then, use natural language processing technology and generative AI methods to extract key disaster characteristics such as the time of occurrence, disaster intensity, impact range, casualties, and property losses of different types of disasters one by one, realize the classification and compilation of massive disaster texts, and create a historical disaster database of rainstorm-flash flood disaster chains. Finally, sort the disaster data in the same area in the order of "occurrence time" to extract all disaster chains. Count the number of disaster chains in the entire basin and identify the hot spots where rainstorm-flash flood disaster chains frequently occur.

[0102] Step 2: Use the method of building a Bayesian network model to identify the triggering threshold of the rainstorm-flash flood disaster chain.

[0103] First, combining expert opinions, literature research, and the historical disaster database created, the influencing factors of the rainstorm-flash flood disaster chain are determined, and the influencing factors, key disaster characteristics, etc. are used as node variables, and the causal relationship between node variables is a directed edge to construct the Bayesian network topology of the rainstorm-flash flood disaster chain. Then, the node variables are discretized to obtain the state value of the node. The prior probability of the node variable is calculated by replacing the probability with the frequency. The EM algorithm is applied for parameter learning to obtain the conditional probability of the node variable. Then, the derivative disaster is used as an evidence variable, and the posterior probability of its parent node is calculated using the Bayesian reasoning method. Finally, in the constructed Bayesian network model, a sufficiently large posterior probability critical value is assumed (according to the extreme event probability classification standard, the posterior probability critical value is set to 90%), and the intensity of the pre-disaster that causes the derivative disaster to occur is reversely calculated, and the intensity threshold of the derivative disaster of flash floods, landslides, and debris flows triggered by heavy rain is obtained.

[0104] In this way, the technical solution proposed in this embodiment can accurately and in real time perform probabilistic reasoning on the rainstorm-flash flood disaster chain, identify the triggering threshold of the rainstorm-flash flood disaster chain, and provide technical support for accurate early warning of the rainstorm-flash flood disaster chain.

[0105] Figure 4 is a schematic diagram of the composition structure of a device for identifying a trigger threshold of a rainstorm-flash flood disaster chain based on a Bayesian network provided by an embodiment of the present invention, such as Figure 4As shown, a Bayesian network-based rainstorm-flash flood disaster chain trigger threshold recognition device 400 includes: an extraction module 401, which is used to extract historical disaster texts related to the rainstorm-flash flood disaster chain from multiple data sources; a processing module 402, which is used to sequentially clean the historical disaster texts according to a preset fuzzy string matching algorithm and a preset text similarity algorithm to obtain processed disaster texts; the extraction module 401 is also used to perform feature extraction on the processed disaster texts to obtain key disaster features; a construction module 403 is used to construct a target Bayesian network topology of the rainstorm-flash flood disaster chain according to the processed disaster texts and the key disaster features; a discretization module 404 is used to extract the historical disaster texts according to the preset fuzzy string matching algorithm and the preset text similarity algorithm to obtain processed disaster texts; the extraction module 401 is also used to extract the features of the processed disaster texts to obtain key disaster features; a construction module 403 is used to construct a target Bayesian network topology of the rainstorm-flash flood disaster chain according to the processed disaster texts and the key disaster features; a discretization module 404 is used to extract the historical disaster texts according to the preset fuzzy string matching algorithm and the preset text similarity algorithm to obtain processed disaster texts ... 04, used to discretize each node variable in the target Bayesian network topology to obtain the state value of the node variable; the state value represents the occurrence and non-occurrence state of the disaster event; the calculation module 405 is used to calculate the prior probability of the node variable, and calculate the conditional probability of the causal relationship between nodes based on the prior probability, the state value and the preset application expectation maximization algorithm; the calculation module 405 is also used to calculate the target posterior probability of the first disaster triggering the derivative disaster; the inversion module 406 is used to invert the intensity of the first disaster according to the preset posterior probability critical value and the target posterior probability, and determine the intensity threshold of the first disaster triggering the derivative disaster.

[0106] It should be noted that the description of the device of the embodiment of the present invention is similar to the description of the above method embodiment, and has similar beneficial effects as the same method embodiment, so it will not be repeated. For technical details not disclosed in the embodiment of the device, please refer to the description of the method embodiment of the present invention for understanding.

[0107] Based on the above embodiments, an embodiment of the present invention further provides an electronic device, Figure 5 FIG. 1 is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Figure 5 As shown, the hardware entity of the electronic device 500 includes: a memory 501 and a processor 502, wherein the memory 501 stores a computer program that can be run on the processor 502, and when the processor 502 executes the program, the steps in the above-mentioned embodiment of the heavy rain-flash flood disaster chain trigger threshold identification based on the Bayesian network are implemented.

[0108] The memory 501 is configured to store instructions and applications executable by the processor 502, and can also cache data to be processed or processed by the processor 502 and various modules in the electronic device 500 (for example, image data, audio data, voice communication data, and video communication data), which can be implemented through flash memory (FLASH) or random access memory (Random Access Memory, RAM).

[0109] Based on the foregoing embodiments, an embodiment of the present invention further provides a computer-readable storage medium. A computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor of an electronic device, it can implement the method for identifying the triggering threshold of the rainstorm-mountain flood disaster chain based on a Bayesian network provided in any previous embodiment.

[0110] The descriptions of the various embodiments above tend to emphasize the differences between the various embodiments. Their similarities or similarities can be referred to each other. For the sake of brevity, they will not be repeated herein.

[0111] The methods disclosed in the method embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments.

[0112] The features disclosed in the product embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new product embodiments.

[0113] The features disclosed in the method or device embodiments provided by the present invention can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0114] It should be noted that the above computer-readable storage medium may be a ferroelectric memory (FRAM, Ferromagnetic Random Access Memory), a read-only memory (ROM, Read Only Memory), a programmable read-only memory (PROM, Programmable Read Only Memory), an erasable programmable read-only memory (EPROM, Erasable Programmable Read Only Memory), an electrically erasable programmable read-only memory (EEPROM, Electrically Erasable Programmable Read Only Memory), a flash memory, a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM, Compact Disk-Read Only Memory), etc.; it may also be various electronic devices including one or any combination of the above memories, such as a mobile phone, a computer, a tablet device, a personal digital assistant, etc.

[0115] It should be noted that in this text, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, such that a process, method or apparatus including a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method or apparatus. Without further limitation, an element defined by the statement "including one..." does not exclude the presence of additional identical elements in the process, method, article or apparatus including such element. In several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of the units is only a logical functional division, and there may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed.

[0116] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0117] Through the description of the above embodiments, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus necessary general hardware nodes. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that makes a contribution to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions for causing a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0118] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing device generate a device for implementing the functions specified in Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0119] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements the functions specified in one process Figure 1 one process or a plurality of processes and / or boxes Figure 1 the functions specified in one box or a plurality of boxes.

[0120] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, such that a series of operational steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one process Figure 1 one process or a plurality of processes and / or boxes Figure 1 the functions specified in one box or a plurality of boxes.

[0121] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for identifying trigger thresholds of rainstorm-flash flood disaster chain based on Bayesian network, characterized in that: The method comprises: Extract historical disaster texts related to the rainstorm-flash flood disaster chain from multiple data sources; According to a preset fuzzy string matching algorithm and a preset text similarity algorithm, the historical disaster information texts are cleaned in sequence to obtain processed disaster information texts; Extracting features from the processed disaster text to obtain key disaster features; Constructing a target Bayesian network topology of the rainstorm-flash flood disaster chain according to the processed disaster text and the key disaster characteristics; Discretize each node variable in the target Bayesian network topology to obtain a state value of the node variable; the state value represents the occurrence or non-occurrence state of the disaster event; Calculating the prior probability of the node variable, and calculating the conditional probability of the causal relationship between the nodes based on the prior probability, the state value and the preset application expectation maximization algorithm; Calculate the target posterior probability that the first disaster triggers the derivative disaster; According to the preset posterior probability critical value and the target posterior probability, the intensity of the preceding disaster is inverted to determine the intensity threshold of the preceding disaster triggering the derivative disaster.

2. The method according to claim 1, characterized in that: The historical disaster information text is cleaned according to a preset fuzzy string matching algorithm and a preset text similarity algorithm to obtain a processed disaster information text, including: According to the preset fuzzy string matching algorithm, the edit distance between two historical disaster texts is calculated; According to the edit distance and the preset text similarity algorithm, the text similarity between the two historical disaster texts is determined; the text similarity is determined by the following formula: ; In the formula, The distance is the minimum number of single-character edit operations required to transform one string into another; and They are the historical disaster texts 1 and 2 to be compared; Represents the length function of a string; Deleting the historical disaster texts whose text similarity exceeds a preset similarity threshold in the historical disaster texts to obtain the processed disaster texts; Feature extraction is performed on the processed disaster text to obtain the key disaster features.

3. The method according to claim 1, characterized in that The target Bayesian network topology of the rainstorm-flash flood disaster chain is constructed according to the processed disaster text and the key disaster characteristics, including: According to the disaster type and the time of occurrence of the disaster, the processed disaster text and the key disaster characteristics are integrated to establish the rainstorm-flash flood disaster chain database; Determining the influencing factors of the rainstorm-flash flood disaster chain according to the rainstorm-flash flood disaster chain database; The target Bayesian network topology is constructed by taking the influencing factors and the key disaster characteristics as node variables and the causal relationships between the node variables as directed edges.

4. A device for identifying the trigger threshold of a rainstorm-flash flood disaster chain based on a Bayesian network, characterized in that: The device comprises: An extraction module is used to extract historical disaster texts related to the rainstorm-flash flood disaster chain from multiple data sources; A processing module, used to perform cleaning processing on the historical disaster information text in sequence according to a preset fuzzy string matching algorithm and a preset text similarity algorithm to obtain a processed disaster information text; The extraction module is further used to extract features from the processed disaster information text to obtain key disaster features; A construction module, used to construct a target Bayesian network topology of the rainstorm-flash flood disaster chain according to the processed disaster text and the key disaster characteristics; A discretization module is used to discretize each node variable in the target Bayesian network topology to obtain a state value of the node variable; the state value represents the occurrence or non-occurrence state of a disaster event; A calculation module, used to calculate the prior probability of the node variable, and calculate the conditional probability of the causal relationship between the nodes according to the prior probability, the state value and the preset application expectation maximization algorithm; The calculation module is also used to calculate the target posterior probability of the first disaster triggering the derivative disaster; The inversion module is used to invert the intensity of the preceding disaster according to a preset posterior probability critical value and the target posterior probability, and determine the intensity threshold of the preceding disaster triggering the derivative disaster.

5. An electronic device, characterized in that: include: A memory for storing executable instructions; The processor is used to implement the Bayesian network-based rainstorm-flash flood disaster chain trigger threshold identification method as described in any one of claims 1 to 3 when executing the executable instructions stored in the memory.

6. A computer-readable storage medium, characterized in that: Executable instructions are stored, which are used to cause the processor to execute the executable instructions to implement the Bayesian network-based rainstorm-flash flood disaster chain trigger threshold identification method described in any one of claims 1 to 3.

Images