A distribution probability calculation method and system for compound flood disasters

Abstract

The present application provides a method and system for calculating the distribution probability of compound flood disasters, which relates to the field of disaster assessment technology. The method includes: constructing a compound disaster knowledge graph based on disasters within a preset time period before and after the occurrence of floods in the study area; extracting chains related to floods in the knowledge graph to construct a compound flood disaster knowledge graph; extracting upstream and downstream disaster events in the knowledge graph; determining the correlation attributes of upstream and downstream disaster events according to the probability of occurrence of upstream and downstream disaster events; obtaining the value of the target disaster-causing factor, determining the value as sample data, and determining the upstream and downstream disaster events with correlation attributes as label-labeled sample data; training a model with labeled sample data to obtain a disaster distribution probability prediction model; inputting the value of the target disaster-causing factor into the disaster distribution probability prediction model for processing, and obtaining the distribution probability of various disasters belonging to indirect related disasters. The purpose is to quantitatively evaluate the probability distribution of compound flood disasters in a specified area.

Description

Technical Field

[0001] The present application relates to the technical field of disaster assessment, and in particular to a distribution probability calculation method and system for compound flood disasters. Background Art

[0002] It is generally believed that compound disasters are multiple disasters occurring simultaneously or multiple disasters following one disaster. Quantitatively assessing the distribution probability of compound flood disasters in a region and improving the construction of disaster prevention facilities in areas with a higher probability of disasters is of great significance for regional precision disaster prevention, mitigation and relief.

[0003] As for the current research on the construction of disaster event knowledge graph, only valid single sentences containing disaster events are considered in text mining (i.e., the extraction of secondary disasters), while concurrent and coupled disaster events that are far away in the text are ignored. Secondly, in the process of constructing the disaster event knowledge graph, it does not include quantitative estimation of the distribution probability of disasters, but only provides a method for constructing the disaster event knowledge graph. Therefore, it is impossible to effectively quantitatively evaluate flood disasters in the region, making it difficult to provide constructive suggestions for the field of disaster assessment and management. Summary of the invention

[0004] In view of this, the present application provides a method and system for calculating the distribution probability of compound flood disasters, aiming to quantitatively evaluate the occurrence probability distribution of compound flood disasters in a specified area, provide a reliable reference for the construction and maintenance of flood control facilities in the specified area, and thus minimize the harm caused by the disaster.

[0005] In a first aspect of the present application, a method for calculating the distribution probability of a compound flood disaster is provided, the method comprising:

[0006] Based on the disaster events that occurred within a preset time period before and after the flood disaster in the study area, a composite disaster knowledge graph is constructed, wherein the nodes of the composite disaster knowledge graph are disaster events and disaster-bearing bodies, and the relationship between the nodes is the start time and the end time;

[0007] Based on the flood corpus, extract the relationship chain related to the flood disaster in the composite disaster knowledge graph, and construct the corresponding composite flood disaster knowledge graph;

[0008] Extract upstream disaster events pointing to flood disasters and downstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph through grammatical statements;

[0009] According to the respective occurrence probabilities of upstream disaster events and downstream disaster events, the correlation attributes of upstream disaster events and downstream disaster events are determined, and the correlation attributes include directly related disasters and indirectly related disasters;

[0010] Obtaining the value of the target disaster factor in the study area, and determining the value as sample data, determining the upstream disaster events and downstream disaster events with correlation attributes in the study area as labels to mark the sample data, and obtaining target sample data;

[0011] The constructed initial disaster distribution probability prediction model is trained and tested by using a preset number of target sample data to obtain a trained disaster distribution probability prediction model;

[0012] The value of the target disaster-causing factor of the area to be predicted is input into the disaster distribution probability prediction model for processing, and the distribution probability of various disasters belonging to indirectly related disasters in the area to be predicted is predicted.

[0013] Optionally, a composite disaster knowledge graph is constructed based on disaster events that occurred within a preset time period before and after the flood disaster in the study area, including:

[0014] By mining the meteorological disaster text data, a disaster event relationship table is constructed;

[0015] Based on the location information in the disaster event relationship table, extracting a first disaster event set occurring in the study area from the disaster event relationship table;

[0016] Determine the flood disaster in the first disaster event set, and extract the second disaster event set that occurs within a preset time period before and after the flood disaster occurs;

[0017] Based on the second disaster event set, a corresponding composite disaster knowledge graph is constructed.

[0018] Optionally, by performing text mining on the meteorological disaster text data, a disaster event relationship table is constructed, including:

[0019] By mining the meteorological disaster text data, we construct the initial disaster event relationship table and disaster event table;

[0020] The initial disaster event relationship table is optimized based on the disaster event table to obtain the disaster event relationship table.

[0021] Optionally, upstream disaster events pointing to flood disasters and downstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph are extracted through grammatical statements, including:

[0022] Extracting upstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph through the first graph database grammar statement;

[0023] The downstream disaster events pointed to by the flood disaster in the composite flood disaster knowledge graph are extracted through the second graph database grammar statement.

[0024] Optionally, identify target hazards, including:

[0025] By using the chi-square test algorithm, the correlation values ​​between various disaster-causing factors and indirect related disasters are determined, wherein the indirect related disasters are disasters indirectly related to flood disasters;

[0026] Compare the correlation value of the disaster-causing factor with the set threshold to obtain a comparison result;

[0027] According to the comparison result, it is determined whether the disaster factor is a target disaster factor.

[0028] Optionally, the constructed initial disaster distribution probability prediction model is trained and tested by using a preset number of target sample data to obtain a trained disaster distribution probability prediction model, including:

[0029] Randomly divide the preset number of target sample data into training sets and test sets;

[0030] Input the target sample data in the training set, and obtain the corresponding distribution probability of various disasters by performing operations at each level of the initial disaster distribution probability prediction model;

[0031] The distribution probability of each type of disaster obtained and the label of the input target sample data are calculated through cross entropy loss to obtain the corresponding loss value;

[0032] Update the weight parameters of the initial disaster distribution probability prediction model through back propagation;

[0033] When the number of iterative trainings of the initial disaster distribution probability prediction model through the training set meets the first condition, the initial disaster distribution probability prediction model is tested and evaluated through the test set to obtain a corresponding evaluation result;

[0034] When the evaluation result indicates that the initial disaster distribution probability prediction model is qualified, obtaining a corresponding disaster distribution probability prediction model;

[0035] When the evaluation result indicates that the initial disaster distribution probability prediction model training is unqualified, a new round of training is performed on the initial disaster distribution probability prediction model using the training set.

[0036] Optionally, before inputting the value of the target disaster factor of the area to be predicted into the disaster distribution probability prediction model for processing and predicting the distribution probability of various disasters belonging to indirectly related disasters in the area to be predicted, the method further includes:

[0037] Based on the correlation values ​​of the target disaster factors, all target disaster factors are sorted from large to small to obtain the sorting results;

[0038] According to the ranking result, the target disaster factors with the smallest correlation value are removed from all target disaster factors in turn to form a new target disaster factor combination;

[0039] Determine the AUC curve value corresponding to each new target disaster factor combination through the disaster distribution probability prediction model;

[0040] The target disaster factor combination corresponding to the maximum AUC curve value is determined as the optimal disaster factor combination;

[0041] The method of inputting the value of the target disaster factor of the area to be predicted into the disaster distribution probability prediction model for processing, and predicting the distribution probability of various disasters belonging to indirectly related disasters in the area to be predicted, includes:

[0042] The values ​​of each disaster factor in the optimal disaster factor combination of the area to be predicted are input into the disaster distribution probability prediction model for processing, and the distribution probability of various disasters belonging to indirectly related disasters in the area to be predicted is predicted.

[0043] Optionally, the chi-square test algorithm is used to determine the correlation values ​​between various disaster-causing factors and indirectly related disasters, including:

[0044] By bringing the data of various disaster-causing factors into the chi-square test algorithm for calculation, the correlation values ​​of various disaster-causing factors and indirect related disasters are obtained. The expression of the chi-square test algorithm is:

[0045] Among them, A is the number of hazard factors observed in various types of indirect related disasters, p refers to the expected probability under the assumption that the hazard factor has no significant difference in various types of indirect related disasters, k is the sum of the number of categories of the hazard factor corresponding to various types of indirect related disasters, It represents the correlation value of the disaster factor, E is the expected value under the assumption that the disaster factor has no significant difference in various types of indirect related disasters, and n is the frequency of occurrence of the disaster factor category.

[0046] Optionally, the method further includes:

[0047] The disaster-prone objects and various disasters belonging to directly related disasters in the area to be predicted are combined with the various disasters belonging to indirectly related disasters obtained by prediction to obtain the distribution probability of various upstream and downstream disasters in the area to be predicted.

[0048] A second aspect of the present application provides a distribution probability calculation system for compound flood disasters, the system comprising:

[0049] The first construction module is used to construct a composite disaster knowledge graph based on disaster events that occurred within a preset time period before and after the flood disaster in the study area, wherein the nodes of the composite disaster knowledge graph are disaster events and disaster-bearing bodies, and the relationship between the nodes is the start time and the end time;

[0050] The second construction module is used to extract the relationship chain related to flood disasters in the composite disaster knowledge graph based on the flood corpus, and construct the corresponding composite flood disaster knowledge graph;

[0051] An extraction module is used to extract upstream disaster events pointing to flood disasters and downstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph through grammatical statements;

[0052] A correlation attribute determination module is used to determine the correlation attributes of the upstream disaster event and the downstream disaster event according to the respective occurrence probabilities of the upstream disaster event and the downstream disaster event, and the correlation attributes include directly related disasters and indirectly related disasters;

[0053] A sample data determination module is used to obtain the value of the target disaster factor in the study area, determine the value as sample data, determine the upstream disaster events and downstream disaster events with correlation attributes in the study area as labels to mark the sample data, and obtain target sample data;

[0054] The model training and testing module is used to train and test the constructed initial disaster distribution probability prediction model through a preset number of target sample data to obtain a qualified disaster distribution probability prediction model;

[0055] The distribution probability prediction module is used to input the value of the target disaster-causing factor of the area to be predicted into the disaster distribution probability prediction model for processing, and predict the distribution probability of various disasters belonging to indirectly related disasters in the area to be predicted.

[0056] Compared with the prior art, this application has the following advantages:

[0057] The present application provides a method for calculating the distribution probability of a composite flood disaster. First, based on the disaster events that occurred within a preset time period before and after the flood disaster in the study area, a composite disaster knowledge graph is constructed. The nodes of the composite disaster knowledge graph are disaster events and disaster-affected entities, and the relationship between the nodes is the start time and the end time; based on the flood corpus, the relationship chain related to the flood disaster in the composite disaster knowledge graph is extracted to construct the corresponding composite flood disaster knowledge graph; the upstream disaster events pointing to the flood disaster and the downstream disaster events pointing to the flood disaster in the composite flood disaster knowledge graph are extracted through grammatical statements; according to the respective probability of occurrence of the upstream disaster events and the downstream disaster events, the correlation attributes of the upstream disaster events and the downstream disaster events are determined. The correlation attributes include directly related disasters and indirectly related disasters; obtain the value of the target disaster factor in the study area, and determine the value as sample data, determine the upstream disaster events and downstream disaster events with correlation attributes in the study area as labels to annotate the sample data, and obtain the target sample data; train and test the constructed initial disaster distribution probability prediction model through a preset number of target sample data, and obtain a trained qualified disaster distribution probability prediction model; input the value of the target disaster factor in the area to be predicted into the disaster distribution probability prediction model for processing, and predict the distribution probability of various disasters belonging to indirect related disasters in the area to be predicted. Therefore, the distribution probability of various disaster events belonging to indirect related disasters in the area to be predicted can be quantitatively evaluated relative to flood disasters in the area to be predicted, thereby providing a reliable reference for the construction and maintenance of flood control facilities in the area to be predicted, thereby minimizing the harm caused by disasters.

[0058] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below.

[0060] Figure 1 A flow chart of a method for calculating the distribution probability of a composite flood disaster provided in an embodiment of the present application;

[0061] Figure 2 A schematic diagram of a composite disaster knowledge graph in a method for calculating the distribution probability of composite flood disasters provided in an embodiment of the present application;

[0062] Figure 3A schematic diagram of a knowledge graph of composite flood disasters in a method for calculating the distribution probability of composite flood disasters provided in an embodiment of the present application;

[0063] Figure 4 A schematic diagram of a downstream disaster event directed by a flood disaster in a distribution probability calculation method of a composite flood disaster provided in an embodiment of the present application;

[0064] Figure 5 A schematic diagram of an upstream disaster event pointing to a flood disaster in a distribution probability calculation method of a composite flood disaster provided in an embodiment of the present application;

[0065] Figure 6 A schematic diagram of the distribution probabilities of various upstream and downstream disasters in a region to be predicted in a method for calculating the distribution probability of a composite flood disaster provided in an embodiment of the present application;

[0066] Figure 7 A schematic diagram of a distribution probability calculation system for a composite flood disaster provided in an embodiment of the present application. DETAILED DESCRIPTION

[0067] Exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings.

[0068] Figure 1 A flow chart of a method for calculating the distribution probability of a composite flood disaster provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method includes:

[0069] Step S1: Based on the disaster events that occurred within a preset time period before and after the flood disaster in the study area, a composite disaster knowledge graph is constructed, wherein the nodes of the composite disaster knowledge graph are disaster events and disaster-affected bodies, and the relationship between the nodes is the start time and the end time.

[0070] In this embodiment, the implementation method for constructing any composite disaster knowledge graph for the study area is the same. Here, the construction of a composite disaster knowledge graph for the study area is used as an example for explanation; for a disaster event in the study area (the disaster event can be any type of disaster event), all disaster events that occurred within a preset time period before and after the disaster event are determined, where all disaster events include disaster-bearing objects of the disaster event, such as products and raw materials, houses and structures, people, economic losses, etc. Then, based on all the determined disaster events, a composite disaster knowledge graph corresponding to the study area is constructed, such as Figure 2 As shown, Figure 2A schematic diagram of two composite disaster knowledge graphs is shown as an example. The node set in the composite disaster knowledge graph consists of all disaster events, one node corresponds to one disaster event, and the relationship between nodes in the composite disaster knowledge graph is the start time and end time of the disaster event.

[0071] Among them, the preset duration is preferably 7 natural days, and the preset time before and after the flood disaster is preferably 7 natural days before the flood disaster and 7 natural days after the flood disaster, a total of 15 natural days.

[0072] In this embodiment, for any disaster event in the study area, the corresponding composite disaster knowledge graph can be constructed through the same implementation method.

[0073] Step S2: Based on the flood corpus, extract the relationship chain related to flood disasters in the composite disaster knowledge graph to construct a corresponding composite flood disaster knowledge graph.

[0074] In this embodiment, the present application pre-constructs a noun corpus representing flood disasters, and the noun corpus is the flood corpus. For a large number of composite disaster knowledge graphs corresponding to the study area constructed in step S1, based on the constructed flood corpus, the flood disaster event nodes matching the flood disaster nouns in the flood corpus are extracted from all composite disaster knowledge graphs, and then the relationship chains related to the flood disaster event nodes are extracted with the flood disaster event nodes as the center. All the extracted relationship chains constitute the corresponding composite flood disaster knowledge graph, such as Figure 3 As shown, Figure 3 A schematic diagram of a composite flood disaster knowledge graph is shown as an example, wherein the relationship between nodes in the composite flood disaster knowledge graph is the start time and end time of the disaster event.

[0075] Step S3: extracting upstream disaster events pointing to flood disasters and downstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph through grammatical statements.

[0076] In this embodiment, in order to facilitate the emergency planning management of compound floods, this application defines the disaster events pointing to floods in the compound flood disaster knowledge graph as upstream disaster events, and the disaster events pointed to by floods are downstream disaster events. For the compound flood disaster knowledge graph corresponding to the research area determined by step S2, the compound flood disaster knowledge graph is extracted through the corresponding NOSQL graph database syntax statement, and the upstream disaster events pointing to floods and the downstream disaster events pointed to by floods in the compound flood disaster knowledge graph can be extracted.

[0077] In the present application, step S3 may include: extracting upstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph through a first graph database grammar statement; extracting downstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph through a second graph database grammar statement.

[0078] In this embodiment, for each composite flood disaster knowledge graph in the study area determined by step S2, the upstream disaster events pointing to the flood disaster in each composite flood disaster knowledge graph are extracted through the first graph database grammar statement "match p = (n)-[*]->(m) where n.name in['keyword'] return p". At the same time, the downstream disaster events pointing to the flood disaster in each composite flood disaster knowledge graph are extracted through the second graph database grammar statement "match p = (m)-[*]->(n) where n.name in['keyword'] return p", such as Figure 4 As shown, Figure 4 The following example shows the downstream disaster events pointed to by flood disasters in the composite flood disaster knowledge graph, such as Figure 5 As shown, Figure 5 An example is shown of upstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph.

[0079] Step S4: According to the respective occurrence probabilities of the upstream disaster event and the downstream disaster event, the correlation attributes of the upstream disaster event and the downstream disaster event are determined, and the correlation attributes include directly related disasters and indirectly related disasters.

[0080] In this embodiment, after determining each upstream disaster event and each downstream disaster event in the composite flood disaster knowledge graph corresponding to the study area through step S3, the occurrence probability of each upstream disaster event and each downstream disaster event in the composite flood disaster knowledge graph is determined. The implementation method of determining the occurrence probability of any upstream disaster event in this application is the same as the implementation method of determining the occurrence probability of any downstream disaster event. Here, an upstream disaster event is used as an example for explanation: the total number of flood disasters occurring in the study area within a certain period of time is counted, and the total number of such upstream disaster events occurring in the study area within a preset period of time before the occurrence of the flood disaster event is counted within the certain period of time. The total number of such upstream disaster events occurring is compared with the total number of flood disasters occurring, and the occurrence probability of such upstream disaster events is obtained, wherein the certain period of time can be set according to the actual application scenario, such as 5 years, 10 years, 20 years, etc.

[0081] For example, it is statistically obtained that X flood disasters occurred in the study area in the past 10 years, and it is statistically obtained that among the X flood disasters that occurred in the study area in the past 10 years, an upstream disaster event a occurred within a preset time before Y flood disasters occurred, so the probability of occurrence of the upstream disaster event a is determined to be Y / X. Based on the same calculation method, each upstream disaster event and each downstream disaster event in the composite flood disaster knowledge graph can determine the corresponding probability of occurrence.

[0082] In this embodiment, after determining the corresponding occurrence probabilities of each upstream disaster event and each downstream disaster event in the composite flood disaster knowledge graph, the corresponding occurrence probabilities of each upstream disaster event and each downstream disaster event in the composite flood disaster knowledge graph are compared with the target threshold. When the corresponding occurrence probability of the upstream disaster event is less than or equal to the target threshold, the correlation attribute of the upstream disaster event is determined to be an indirect related disaster, that is, the upstream disaster event is indirectly related to the flood disaster; when the corresponding occurrence probability of the upstream disaster event is greater than the target threshold, the correlation attribute of the upstream disaster event is determined to be a direct related disaster, that is, the upstream disaster event is directly related to the flood disaster; when the corresponding occurrence probability of the downstream disaster event is less than or equal to the target threshold, the correlation attribute of the downstream disaster event is determined to be an indirect related disaster, that is, the downstream disaster event is indirectly related to the flood disaster; when the corresponding occurrence probability of the downstream disaster event is greater than the target threshold, the correlation attribute of the downstream disaster event is determined to be a direct related disaster, that is, the downstream disaster event is directly related to the flood disaster. Among them, the target threshold is preferably 80%. Thus, each upstream disaster event and each downstream disaster event in all the composite flood disaster knowledge graphs in the study area will be determined, and at the same time, the specific correlation attributes of each upstream disaster event and the correlation attributes of each downstream disaster event will be determined. Accordingly, for all the composite flood disaster knowledge graphs in the study area, each upstream disaster event and each downstream disaster event in each composite flood disaster knowledge graph can be determined by the same implementation as step S3, and at the same time, each specific correlation attribute of each upstream disaster event and the correlation attribute of each downstream disaster event can be determined by the same implementation as step S4.

[0083] Step S5: Obtain the value of the target disaster factor in the study area, and determine the value as sample data, determine the upstream disaster events and downstream disaster events with correlation attributes in the study area as labels to mark the sample data, and obtain target sample data.

[0084] In this embodiment, the construction method for any target sample data is the same, and the construction of a target sample data is used as an example for explanation. After obtaining the upstream disaster events pointing to the flood disaster and the downstream disaster events pointing to the flood disaster in the composite flood disaster knowledge graph of the study area through the above steps S1 to S4, and determining the correlation attributes of the upstream disaster events and the correlation attributes of the downstream disaster events, the upstream disaster events with the correlation attribute of indirect related disasters in the upstream disaster events and the downstream disaster events with the correlation attribute of indirect related disasters in the downstream disaster events are determined as a label, and the label will record the upstream disaster events with the correlation attribute of indirect related disasters and the correlation attribute of the upstream disaster events, as well as the downstream disaster events with the correlation attribute of indirect related disasters and the correlation attribute of the downstream disaster events. Then, the value of the target disaster factor in the study area is determined as sample data, and the sample data is labeled with the label to obtain a corresponding target sample data.

[0085] For example, through the above steps S1 to S4, the upstream disaster events U1, U2, U3 pointing to flood disasters and the downstream disaster events D1, D2, D3, D4 pointing to flood disasters in a composite flood disaster knowledge graph of a research area are obtained, and the correlation attributes of the upstream disaster events are determined to be U1 as an indirect related disaster, U2 as an indirect related disaster, and U3 as a direct related disaster, and the correlation attributes of the downstream disaster events are determined to be D1 as a direct related disaster, D2 as an indirect related disaster, D3 as an indirect related disaster, and D4 as a direct related disaster. Then, the upstream disaster events U1 and U2 with the correlation attribute of indirect related disasters in the upstream disaster events, and the downstream disaster events D2 and D3 with the correlation attribute of indirect related disasters in the downstream disaster events are jointly determined as a label for a sample data, and the label will record the upstream disaster events U1 and U2 and the correlation attributes of the two disaster events, and record the downstream disaster events D2 and D3 and the correlation attributes of the two disaster events. Finally, the value f1 of the target disaster factor F1, the value f2 of the target disaster factor F2, and the value f3 of the target disaster factor F3 in the study area are collectively determined as a sample data, and the sample data is annotated with the label to obtain a target sample data.

[0086] Step S6: Train and test the constructed initial disaster distribution probability prediction model using a preset number of target sample data to obtain a qualified disaster distribution probability prediction model.

[0087] In this embodiment, the present application can construct corresponding target sample data for each study area through the implementation of the above steps S1 to S5. At the same time, when multiple flood disaster events have occurred in the same study area in the past, a corresponding composite flood disaster knowledge graph can be obtained for any flood disaster event. Based on the composite flood disaster knowledge graph, a corresponding target sample data can be constructed through the implementation of the above steps S1 to S5, that is, the same study area can also have multiple target sample data. The present application pre-constructs a preset number of target sample data through the implementation of the above steps S1 to S5. The preset number of target sample data involves target sample data of multiple study areas. The preset number can be set to various values ​​according to the actual application scenario, and is not specifically limited here. This application also pre-builds an untrained initial disaster distribution probability prediction model. After the initial disaster distribution probability prediction model is trained and qualified, it will be used to predict the distribution probability of disaster events in the predicted area. The initial disaster distribution probability prediction model is preferably a multi-layer perception deep learning model (MLP). It should be understood that this is only the preferred model of the initial disaster distribution probability prediction model. The initial disaster distribution probability prediction model can also be other deep learning models. Then, the initial disaster distribution probability prediction model is trained and tested with a preset number of target sample data, and the corresponding trained and qualified disaster distribution probability prediction model is obtained after the test passes.

[0088] Step S7: Input the value of the target disaster factor of the area to be predicted into the disaster distribution probability prediction model for processing, and predict the distribution probability of various disasters belonging to indirectly related disasters in the area to be predicted.

[0089] In this embodiment, after obtaining a qualified disaster distribution probability prediction model through training in step S6, the values ​​of various target disaster factors in the area to be predicted are input into the qualified disaster distribution probability prediction model to perform prediction processing on the distribution probability of disaster events, thereby obtaining corresponding output results. The output results include various types of indirect related disaster events that may occur in the area to be predicted, whether various types of indirect related disaster events belong to upstream disaster events or downstream disaster events, and the distribution probability of various types of indirect related disaster events, that is, the probability of occurrence.

[0090] Therefore, the present application provides a method for calculating the distribution probability of a composite flood disaster. First, based on the disaster events that occurred within a preset time period before and after the flood disaster in the study area, a composite disaster knowledge graph is constructed. The nodes of the composite disaster knowledge graph are disaster events and disaster-bearing bodies, and the relationship between the nodes is the start time and the end time; based on the flood corpus, the relationship chain related to the flood disaster in the composite disaster knowledge graph is extracted to construct the corresponding composite flood disaster knowledge graph; the upstream disaster events pointing to the flood disaster and the downstream disaster events pointed to by the flood disaster in the composite flood disaster knowledge graph are extracted through grammatical statements; the upstream disaster events are determined according to the respective probability of occurrence of the upstream disaster events and the downstream disaster events. The correlation attributes of the upstream and downstream disaster events are obtained, and the correlation attributes include directly related disasters and indirectly related disasters; the values ​​of the target disaster factors in the study area are obtained, and the values ​​are determined as sample data, and the upstream disaster events and downstream disaster events with correlation attributes in the study area are determined as labels to annotate the sample data to obtain the target sample data; the constructed initial disaster distribution probability prediction model is trained and tested through a preset number of target sample data to obtain a trained qualified disaster distribution probability prediction model; the values ​​of the target disaster factors in the area to be predicted are input into the disaster distribution probability prediction model for processing, and the distribution probability of various disasters belonging to indirect related disasters in the area to be predicted is predicted. Therefore, the distribution probability of various disaster events belonging to indirect related disasters in the area to be predicted can be quantitatively evaluated relative to flood disasters in the area to be predicted, thereby providing a reliable reference for the construction and maintenance of flood control facilities in the area to be predicted, thereby minimizing the harm caused by disasters.

[0091] In combination with the above embodiments, in one implementation, the present application embodiment further provides a method for calculating the distribution probability of a compound flood disaster. In the method for calculating the distribution probability of a compound flood disaster, step S1 may include steps S11 to S14:

[0092] Step S11: Construct a disaster event relationship table by performing text mining on meteorological disaster text data.

[0093] In this embodiment, by performing text mining on the text data of the meteorological disaster bulletin, various disaster events that occurred in the past, as well as the correlation between the locations of various disaster events and various disaster events, and the start time and end time of each disaster event, are determined, and a disaster event relationship table is constructed based on the various types of data information obtained. Among them, each row of data in the disaster event relationship table records two disaster events (or one disaster event and one disaster-bearing body, or two disaster-bearing bodies) that have a correlation, as well as the earliest start time of the two occurrence times and the latest end time of the two end times of the two disaster events, and the location of the two disaster events. As shown in Table 1, an example of a disaster event relationship table is shown, in which only the disaster events of city A with the start and end times of t1 and t2 respectively are shown, and the disaster event relationship table actually constructed will record disaster events of different cities and different start and end events. For ease of understanding and description, this application also describes the disaster-bearing body as a disaster event.

[0094] Table 1

[0095]

[0096] Step S12: Based on the location information in the disaster event relationship table, extract a first disaster event set occurring in the study area from the disaster event relationship table.

[0097] In this embodiment, after constructing a disaster event relationship table including a large number of cities through step S11, based on the location information recorded in each row of data in the disaster event relationship table, the rows of data occurring in the study area are extracted to form a first disaster event set.

[0098] Step S13: Determine the flood disaster in the first disaster event set, and extract the second disaster event set that occurs within a preset time period before and after the flood disaster.

[0099] In this embodiment, after extracting the rows of data occurring in the study area from the disaster event relationship table to form the first disaster event set through step S12, the start time of the flood disaster event in the first disaster event set and the data row corresponding to the flood disaster event is determined, and the start time is determined as the occurrence time of the flood disaster event. Or, based on the text mining of the meteorological disaster bulletin text data, a disaster event table is constructed, and the event type, occurrence time, occurrence location, disaster scale and other information of each disaster event obtained by mining are recorded in the disaster event table, and then based on the determined flood disaster event in the first disaster event set, the occurrence time of the flood disaster event is determined from the disaster event table. Based on the determined occurrence time, all data rows within the preset time length before and after the occurrence of the flood disaster event are extracted from the first disaster event set, and all data rows constitute the second disaster event set.

[0100] Step S14: Based on the second disaster event set, construct a corresponding composite disaster knowledge graph.

[0101] In this embodiment, after determining to obtain a second disaster event set corresponding to a flood disaster event through step S13, a corresponding composite disaster knowledge graph is constructed based on the second disaster event set. Specifically, based on the obtained second disaster event set, the second disaster event set is processed by neo4j software to obtain a corresponding composite disaster knowledge graph.

[0102] In combination with the above embodiments, in one implementation, the present application embodiment further provides a method for calculating the distribution probability of a compound flood disaster. In the method for calculating the distribution probability of a compound flood disaster, step S11 may include steps S111 to S112:

[0103] Step S111: construct an initial disaster event relationship table and a disaster event table by performing text mining on the meteorological disaster text data.

[0104] In this embodiment, text mining is performed on the text data of meteorological disaster bulletins to determine various disaster events that have occurred in the past, and a disaster event table is constructed based on the various disaster events that have occurred in the past. The disaster event table will record information such as the event type, occurrence time, location, and scale of disaster of each disaster event obtained by mining. At the same time, by performing text mining on the meteorological disaster bulletin text data, the locations of various disaster events recorded in the meteorological disaster bulletin text data and the influence relationship between various disaster events, as well as the start time and end time of each disaster event are determined, and then based on the determined disaster events that have occurred in the past, and the locations of various disaster events recorded in the meteorological disaster bulletin text data obtained by mining and the influence relationship between various disaster events, as well as the start time and end time of each disaster event, a corresponding initial disaster event relationship table is constructed. The initial disaster event relationship table is directly obtained based on the meteorological disaster text data, as shown in Table 2 below. Table 2 is an example of a disaster event table. This example simply shows some disaster events that occurred in a certain period of time in city A, while the actual disaster event table will involve disaster events in a large number of cities and a large number of time periods; as shown in Table 3 below, Table 3 is an example of an initial disaster event relationship table. This example simply shows the influence relationship between some disaster events that occurred in a certain period of time in city A, while the actual disaster event table will involve the influence relationship between disaster events in a large number of cities and a large number of time periods.

[0105] Table 2

[0106]

[0107] Table 3

[0108]

[0109] Step S112: Optimize the initial disaster event relationship table based on the disaster event table to obtain a disaster event relationship table.

[0110] In this embodiment, since this application focuses more on the temporal relationship for compound disaster events. Therefore, for multiple disaster events that occur within the preset time before and after the flood disaster, although some disaster events do not have an influence relationship with each other in the meteorological disaster bulletin text data, they all occur within the preset time before and after the flood disaster, so it is determined that these disaster events are also part of the compound disaster event. Therefore, for this part of the disaster events, when creating a compound disaster event knowledge graph corresponding to a flood disaster event, the multiple disaster events that occurred within the preset time before and after the flood disaster will be determined based on the created disaster event table, and combined with the data recorded in the initial disaster event relationship table, it is determined which disaster events are correlated but do not have an influence relationship in the initial disaster event relationship table. At this time, the correlation between these disaster events is created and updated to the initial disaster event relationship table. Specifically, the occurrence time of every two disaster events in the initial disaster event relationship table before updating is compared. When the occurrence time interval between the two disaster events does not exceed the set threshold, and the two disaster events do not form a row of data in the initial disaster event relationship table, the two disaster events are filled into the initial disaster event relationship table to form a new row of data. Through the same implementation method, after every two disaster events that meet the conditions in the initial disaster event relationship table are filled into the initial disaster event relationship table, since this application pays more attention to the relationship in time, the impact relationship recorded in the updated initial disaster event relationship table will be removed, and then the corresponding disaster event relationship table is generated based on the updated initial disaster event relationship table with the impact relationship removed. As shown in Table 4 below, Table 4 shows an example of an updated initial disaster event relationship table obtained after the initial disaster event relationship table in Table 3 is updated; as shown in Table 1 above, Table 1 shows the final disaster time relationship table obtained after the impact relationship is removed for the updated initial disaster event relationship table in Table 4.

[0111] Table 4

[0112]

[0113] In combination with the above embodiments, in one implementation, the present application embodiment further provides a method for calculating the distribution probability of a compound flood disaster. In the method for calculating the distribution probability of a compound flood disaster, determining the target disaster factor includes steps S01 to S03:

[0114] Step S01: using a chi-square test algorithm, determining the correlation values ​​between various disaster-causing factors and indirectly related disasters, wherein the indirectly related disasters are disasters indirectly related to flood disasters.

[0115] In this embodiment, there are many influencing factors related to the occurrence of disaster events. Using all influencing factors for the distribution probability prediction of natural disasters will result in low prediction efficiency. Therefore, this application first determines various target disaster factors that will be used for the distribution probability prediction of natural disasters from a large number of various disaster factors. Specifically: through expert evaluation and reference to relevant literature, obtain disaster factors that are correlated with various disasters indirectly related to flood disasters from the two perspectives of natural environment and social environment. Then, through the chi-square test algorithm, calculate the correlation values ​​between various disaster factors with correlation and indirectly related disasters. The higher the correlation value of a disaster factor, the higher the correlation between the disaster factor and the directly related disaster.

[0116] Step S02: Compare the correlation value of the disaster factor with the set threshold to obtain a comparison result.

[0117] In this embodiment, after the correlation values ​​of various disaster-causing factors are determined in step S01, the correlation values ​​of various disaster-causing factors are compared with the set thresholds to obtain the comparison results corresponding to the various disaster-causing factors. The set thresholds can be set according to the actual application scenario and are not limited here.

[0118] Step S03: Determine whether the disaster factor is a target disaster factor according to the comparison result.

[0119] In this embodiment, based on the obtained comparison results corresponding to various disaster factors, when the comparison result corresponding to a disaster factor is that the correlation value of the disaster factor is greater than or equal to the set threshold, the disaster factor is determined as the target disaster factor. When the comparison result corresponding to a disaster factor is that the correlation value of the disaster factor is less than the set threshold, the disaster factor will be discarded and will not participate in the prediction of the subsequent disaster distribution probability.

[0120] In combination with the above embodiments, in one implementation, the present application embodiment further provides a method for calculating the distribution probability of a compound flood disaster. In the method for calculating the distribution probability of a compound flood disaster, step S6 may include steps S61 to S67:

[0121] Step S61: randomly divide a preset number of target sample data into training sets and test sets.

[0122] In this embodiment, a preset number of target sample data sets are constructed and randomly divided into a training set and a test set according to a preset allocation ratio, wherein the preset allocation ratio is preferably 80% and 20%, that is, 80% of the preset number of target sample data are determined as the training set, and 20% of the preset number of target sample data are determined as the test set.

[0123] Step S62: Input the target sample data in the training set, and obtain the corresponding distribution probability of each type of disaster by performing operations at each level of the initial disaster distribution probability prediction model.

[0124] In this embodiment, the initial disaster distribution probability prediction model constructed by inputting the target sample data in the training set is used for model training, and the operations are performed through each level of the initial disaster distribution probability prediction model to obtain the output results of the model output corresponding to the input target sample data, and the output results include the probability of occurrence of various types of disasters, and whether the various types of disasters belong to upstream disasters or downstream disasters, and the correlation attributes of the various types of disasters. The initial disaster distribution probability prediction model predicts disasters that are indirectly related to flood disasters, so the correlation attributes of the various types of disasters output are all indirectly related disasters.

[0125] Step S63: Calculate the distribution probabilities of various disasters and the labels of the input target sample data through cross entropy loss to obtain corresponding loss values.

[0126] In this embodiment, the present application calculates the obtained output result and the label of the input target sample data through the cross entropy loss function to obtain the corresponding loss value.

[0127] Step S64: Update the weight parameters of the initial disaster distribution probability prediction model through back propagation.

[0128] In this embodiment, based on the obtained loss value, the weight parameters of the initial disaster distribution probability prediction model are updated by back propagation to obtain an updated initial disaster distribution probability prediction model, and then the process returns to step S62 to perform steps S62 to S64 in sequence until the preset number of times is returned, and then step S65 is executed. The preset number of times can be set according to the actual application scenario, and is not limited here, such as 50 times, 70 times, etc.

[0129] Step S65: When the number of iterative trainings of the initial disaster distribution probability prediction model through the training set meets the first condition, the initial disaster distribution probability prediction model is tested and evaluated through the test set to obtain a corresponding evaluation result.

[0130] In this embodiment, when the number of iterative training of the initial disaster distribution probability prediction model through the training set reaches the preset number of times, the number of iterative training meets the first condition, and the trained initial disaster distribution probability prediction model is tested and evaluated through the test set to obtain the corresponding evaluation result. The evaluation types include but are not limited to evaluation from accuracy, precision and recall, F1-score, ROC curve, etc.

[0131] Step S66: when the evaluation result indicates that the initial disaster distribution probability prediction model is qualified, obtaining the corresponding disaster distribution probability prediction model.

[0132] In this embodiment, the application sets corresponding qualification standards for different evaluation types. When it is determined that the evaluation results meet the set qualification standards, it is determined that the initial disaster distribution probability prediction model training is qualified, and the corresponding disaster distribution probability prediction model is obtained.

[0133] Step S67: When the evaluation result indicates that the initial disaster distribution probability prediction model training is unqualified, a new round of training is performed on the initial disaster distribution probability prediction model using the training set.

[0134] In this embodiment, when it is determined that the evaluation result meets the set qualification standard, it is determined that the initial disaster distribution probability prediction model training is unqualified. At this time, the system returns to step S61 to execute steps S61 to S67 in sequence until the evaluation result meets the set qualification standard.

[0135] In combination with the above embodiments, in one implementation, the present application embodiment further provides a method for calculating the distribution probability of a compound flood disaster. In the method for calculating the distribution probability of a compound flood disaster, before step S7, the method further includes steps S701 to S704:

[0136] Step S701: based on the correlation values ​​of the target disaster factors, all target disaster factors are sorted in descending order to obtain a sorting result.

[0137] In this embodiment, based on the correlation value of each target disaster factor, all target disaster factors are sorted in descending order to obtain a corresponding sorting result.

[0138] Step S702: according to the ranking result, the target disaster factors with the smallest correlation value are removed from all the target disaster factors in turn to form a new target disaster factor combination.

[0139] In this embodiment, in order to avoid excessive consumption of computing resources caused by considering too many disaster factors, the present application eliminates some disaster factors while ensuring the effect. Specifically: based on the sorting result of the target disaster factors obtained by step S701, each time a target disaster factor with the smallest correlation value is eliminated from all the target disaster factors in the sorting result, and then the remaining target disaster factors are determined as a new target disaster factor combination, and the subsequent step S703 will be executed for each new target disaster factor combination obtained. For example, the sorting result includes target disaster factors A1, A2, A3, A4, and A5 sorted from large to small. First, remove the target disaster factor A5 with the smallest correlation value to form a new target disaster factor combination, which includes target disaster factors A1, A2, A3, and A4; at this time, A5 has been removed, and then when removing the target disaster factor with the smallest correlation value from it in turn, A4 is removed, and at this time a new target disaster factor combination is formed, which includes target disaster factors A1, A2, and A3; at this time, A4 has been removed, and then when removing the target disaster factor with the smallest correlation value from it in turn, A3 is removed, and at this time a new target disaster factor combination is formed, which includes target disaster factors A1 and A2; at this time, A3 has been removed, and then when removing the target disaster factor with the smallest correlation value from it in turn, A2 is removed, and at this time a new target disaster factor combination is formed, which includes target disaster factor A1.

[0140] Step S703: Determine the AUC curve value corresponding to each new target disaster factor combination through the disaster distribution probability prediction model.

[0141] In this embodiment, each new target disaster factor combination is input into the disaster distribution probability prediction model for processing, and the AUC curve value corresponding to each new target disaster factor combination is obtained.

[0142] Continuing with the example in the above step S702, each new target disaster factor combination is input into the disaster distribution probability prediction model for processing, and the AUC curve values ​​corresponding to the target disaster factor combinations A1, A2, A3, A4, and A5 will be obtained, as well as the AUC curve values ​​corresponding to the target disaster factor combinations A1, A2, A3, and A4 will be obtained, as well as the AUC curve values ​​corresponding to the target disaster factor combinations A1 and A2 will be obtained, and the AUC curve values ​​corresponding to the target disaster factor combinations A1 and A2 will be obtained.

[0143] Step S704: Determine the target disaster factor combination corresponding to the maximum AUC curve value as the optimal disaster factor combination.

[0144] In this embodiment, for each new target disaster factor combination corresponding to the AUC curve value obtained in step S703, the target disaster factor combination corresponding to the largest AUC curve value is selected as the optimal disaster factor combination.

[0145] Continuing with the example in the above step S703, by comparing the AUC curve values ​​corresponding to each new target disaster factor combination, it is determined that the AUC curve value corresponding to the target disaster factor combination A1, A2, A3 has the largest AUC curve value, and therefore the target disaster factor combination A1, A2, A3 is determined as the optimal disaster factor combination.

[0146] In the present application, when the method also includes steps S701 to S704, step S7 may include: inputting the values ​​of each disaster factor in the optimal disaster factor combination of the area to be predicted into the disaster distribution probability prediction model for processing, and predicting the distribution probability of various types of disasters that are indirectly related disasters in the area to be predicted.

[0147] In this embodiment, the values ​​of each target disaster factor in the determined optimal disaster factor combination are input into the disaster distribution probability prediction model for processing to obtain the corresponding output results, which include various types of indirect related disaster events that may occur in the area to be predicted, and whether various types of indirect related disaster events belong to upstream disaster events or downstream disaster events, and the distribution probability of various types of indirect related disaster events, that is, the probability of occurrence.

[0148] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a method for calculating the distribution probability of a compound flood disaster. In the method for calculating the distribution probability of a compound flood disaster, step S01 may include: by bringing the data of various disaster-causing factors into the chi-square test algorithm for calculation, respectively, to obtain the correlation values ​​of various disaster-causing factors and indirect related disasters, and the expression of the chi-square test algorithm is:

[0149] Among them, A is the number of hazard factors observed in various types of indirect related disasters, p refers to the expected probability under the assumption that the hazard factor has no significant difference in various types of indirect related disasters, k is the sum of the number of categories of the hazard factor corresponding to various types of indirect related disasters, It represents the correlation value of the disaster factor, E is the expected value under the assumption that the disaster factor has no significant difference in various types of indirect related disasters, and n is the frequency of occurrence of the disaster factor category.

[0150] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a method for calculating the distribution probability of a compound flood disaster. In the method for calculating the distribution probability of a compound flood disaster, the method further includes: merging the disaster-bearing body and various disasters belonging to directly related disasters in the area to be predicted with the various disasters belonging to indirectly related disasters obtained by prediction, and obtaining the distribution probability of various upstream and downstream disasters in the area to be predicted.

[0151] In this embodiment, the various disasters predicted to be indirectly related disasters are combined with the disaster-bearing bodies in the area to be predicted to obtain the distribution probability of various upstream and downstream disasters in the area to be predicted, such as Figure 6 shown.

[0152] Based on the same inventive concept, the present application provides a distribution probability calculation system 700 for a composite flood disaster, such as Figure 7 As shown, the composite flood disaster distribution probability calculation system 700 includes:

[0153] The first construction module 701 is used to construct a composite disaster knowledge graph based on disaster events that occurred within a preset time period before and after the flood disaster in the study area, wherein the nodes of the composite disaster knowledge graph are disaster events and disaster-bearing bodies, and the relationship between the nodes is the start time and the end time;

[0154] The second construction module 702 is used to extract the relationship chain related to flood disasters in the composite disaster knowledge graph based on the flood corpus, and construct the corresponding composite flood disaster knowledge graph;

[0155] An extraction module 703 is used to extract upstream disaster events pointing to flood disasters and downstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph through grammatical statements;

[0156] A correlation attribute determination module 704 is used to determine the correlation attributes of the upstream disaster event and the downstream disaster event according to the respective occurrence probabilities of the upstream disaster event and the downstream disaster event, and the correlation attributes include directly related disasters and indirectly related disasters;

[0157] The sample data determination module 705 is used to obtain the value of the target disaster factor in the study area, determine the value as sample data, determine the upstream disaster events and downstream disaster events with correlation attributes in the study area as labels to mark the sample data, and obtain target sample data;

[0158] The model training and testing module 706 is used to train and test the constructed initial disaster distribution probability prediction model through a preset number of target sample data to obtain a trained disaster distribution probability prediction model;

[0159] The distribution probability prediction module 707 is used to input the value of the target disaster-causing factor of the area to be predicted into the disaster distribution probability prediction model for processing, and predict the distribution probability of various disasters belonging to indirectly related disasters in the area to be predicted.

[0160] Optionally, the first building module 701 includes:

[0161] A disaster event relationship table construction module is used to construct a disaster event relationship table by performing text mining on meteorological disaster text data;

[0162] A first disaster event set determination module, configured to extract a first disaster event set occurring in a study area from the disaster event relationship table based on location information in the disaster event relationship table;

[0163] A second disaster event set determination module, used to determine the flood disaster in the first disaster event set, and extract the second disaster event set that occurs within a preset time period before and after the flood disaster;

[0164] The first construction submodule is used to construct a corresponding composite disaster knowledge graph based on the second disaster event set.

[0165] Optionally, the disaster event relationship table construction module includes:

[0166] The disaster event relationship table construction submodule is used to construct the initial disaster event relationship table and disaster event table by performing text mining on meteorological disaster text data;

[0167] The optimization module is used to optimize the initial disaster event relationship table based on the disaster event table to obtain the disaster event relationship table.

[0168] Optionally, the extraction module 703 includes:

[0169] A first extraction module is used to extract upstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph through a first graph database grammar statement;

[0170] The second extraction module is used to extract downstream disaster events pointed to by flood disasters in the composite flood disaster knowledge graph through a second graphic database grammar statement.

[0171] Optionally, a target disaster factor determination module is used to determine a target disaster factor; the target disaster factor determination module includes:

[0172] A correlation determination module is used to determine the correlation values ​​of various disaster-causing factors and indirect related disasters through a chi-square test algorithm, wherein the indirect related disasters are disasters indirectly related to flood disasters;

[0173] A comparison module is used to compare the correlation value of the disaster-causing factor with the set threshold value to obtain a comparison result;

[0174] The target disaster factor determination submodule is used to determine whether the disaster factor is a target disaster factor according to the comparison result.

[0175] Optionally, the model training and testing module 706 includes:

[0176] A data set partitioning module is used to randomly partition a preset number of target sample data into training sets and test sets;

[0177] The distribution probability determination module is used to input the target sample data in the training set and obtain the corresponding distribution probability of various disasters by performing operations at each level of the initial disaster distribution probability prediction model;

[0178] The loss determination module is used to calculate the distribution probability of various disasters and the labels of the input target sample data through cross entropy loss to obtain the corresponding loss value;

[0179] A parameter updating module is used to update the weight parameters of the initial disaster distribution probability prediction model through back propagation;

[0180] An evaluation result determination module is used to test and evaluate the initial disaster distribution probability prediction model through a test set to obtain a corresponding evaluation result when the number of iterative training of the initial disaster distribution probability prediction model through a training set meets the first condition;

[0181] A disaster distribution probability prediction model determination module is used to obtain a corresponding disaster distribution probability prediction model when the evaluation result indicates that the initial disaster distribution probability prediction model is qualified;

[0182] The training result determination module is used to perform a new round of training on the initial disaster distribution probability prediction model through the training set when the evaluation result indicates that the initial disaster distribution probability prediction model training is unqualified.

[0183] Optionally, the composite flood disaster distribution probability calculation system 700 further includes:

[0184] A sorting module is used to sort all target disaster factors in descending order based on the correlation values ​​of the target disaster factors to obtain a sorting result;

[0185] A target disaster factor combination determination module is used to remove the target disaster factor with the smallest correlation value from all target disaster factors in turn according to the sorting result to form a new target disaster factor combination;

[0186] An AUC curve value determination module is used to determine the AUC curve value corresponding to each new target disaster factor combination through the disaster distribution probability prediction model;

[0187] The optimal disaster factor combination determination module is used to determine the target disaster factor combination corresponding to the maximum AUC curve value as the optimal disaster factor combination;

[0188] The distribution probability prediction module 707 is used to input the values ​​of each disaster factor in the optimal disaster factor combination of the area to be predicted into the disaster distribution probability prediction model for processing, and predict the distribution probability of various disasters belonging to indirectly related disasters in the area to be predicted.

[0189] Optionally, the correlation determination module is used to obtain the correlation values ​​of various disaster-causing factors and indirectly related disasters by respectively bringing the data of various disaster-causing factors into the chi-square test algorithm for calculation. The expression of the chi-square test algorithm is:

[0190] Among them, A is the number of hazard factors observed in various types of indirect related disasters, p refers to the expected probability under the assumption that the hazard factor has no significant difference in various types of indirect related disasters, k is the sum of the number of categories of the hazard factor corresponding to various types of indirect related disasters, It represents the correlation value of the disaster factor, E is the expected value under the assumption that the disaster factor has no significant difference in various types of indirect related disasters, and n is the frequency of occurrence of the disaster factor category.

[0191] Optionally, the composite flood disaster distribution probability calculation system 700 further includes:

[0192] The distribution probability determination module is used to merge the disaster-prone bodies and various disasters belonging to directly related disasters in the area to be predicted with the various disasters belonging to indirectly related disasters obtained by prediction, so as to obtain the distribution probability of various upstream and downstream disasters in the area to be predicted.

[0193] As for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0194] It should be noted that, for the method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited by the described order of actions, because according to the embodiments of the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.

[0195] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0196] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the embodiments of the present application may adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the embodiments of the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0197] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0198] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0199] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0200] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0201] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0202] The above is a detailed introduction to the distribution probability calculation method and system for a composite flood disaster provided by the present application. This article uses specific examples to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for calculating the distribution probability of compound flood disasters, characterized in that: The method comprises: Based on the disaster events that occurred within a preset time period before and after the flood disaster in the study area, a composite disaster knowledge graph is constructed, wherein the nodes of the composite disaster knowledge graph are disaster events and disaster-bearing bodies, and the relationship between the nodes is the start time and the end time; Based on the flood corpus, extract the relationship chain related to the flood disaster in the composite disaster knowledge graph, and construct the corresponding composite flood disaster knowledge graph; Extract upstream disaster events pointing to flood disasters and downstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph through grammatical statements; According to the respective occurrence probabilities of upstream disaster events and downstream disaster events, the correlation attributes of upstream disaster events and downstream disaster events are determined, and the correlation attributes include directly related disasters and indirectly related disasters; Obtaining the value of the target disaster factor in the study area, and determining the value as sample data, determining the upstream disaster events and downstream disaster events with correlation attributes in the study area as labels to mark the sample data, and obtaining target sample data; The constructed initial disaster distribution probability prediction model is trained and tested by using a preset number of target sample data to obtain a trained disaster distribution probability prediction model; The value of the target disaster-causing factor of the area to be predicted is input into the disaster distribution probability prediction model for processing, and the distribution probability of various disasters belonging to indirectly related disasters in the area to be predicted is predicted.

2. The method for calculating the distribution probability of a composite flood disaster according to claim 1, characterized in that: Based on the disaster events that occurred within a preset time before and after the flood disaster in the study area, a composite disaster knowledge graph is constructed, including: By mining the meteorological disaster text data, a disaster event relationship table is constructed; Based on the location information in the disaster event relationship table, extracting a first disaster event set occurring in the study area from the disaster event relationship table; Determine the flood disaster in the first disaster event set, and extract the second disaster event set that occurs within a preset time period before and after the flood disaster occurs; Based on the second disaster event set, a corresponding composite disaster knowledge graph is constructed.

3. The method for calculating the distribution probability of a composite flood disaster according to claim 2, characterized in that: By mining the meteorological disaster text data, a disaster event relationship table is constructed, including: By mining the meteorological disaster text data, we construct the initial disaster event relationship table and disaster event table; The initial disaster event relationship table is optimized based on the disaster event table to obtain the disaster event relationship table.

4. The method for calculating the distribution probability of a composite flood disaster according to claim 1, characterized in that: The upstream disaster events pointing to flood disasters and the downstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph are extracted through grammatical statements, including: Extracting upstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph through the first graph database grammar statement; The downstream disaster events pointed to by the flood disaster in the composite flood disaster knowledge graph are extracted through the second graph database grammar statement.

5. The method for calculating the distribution probability of a composite flood disaster according to claim 1, characterized in that: Identify target hazards, including: By using the chi-square test algorithm, the correlation values ​​between various disaster-causing factors and indirect related disasters are determined, wherein the indirect related disasters are disasters indirectly related to flood disasters; Compare the correlation value of the disaster-causing factor with the set threshold to obtain a comparison result; According to the comparison result, it is determined whether the disaster factor is a target disaster factor.

6. The method for calculating the distribution probability of a composite flood disaster according to claim 1, characterized in that: The constructed initial disaster distribution probability prediction model is trained and tested through a preset number of target sample data to obtain a qualified disaster distribution probability prediction model, including: Randomly divide the preset number of target sample data into training sets and test sets; Input the target sample data in the training set, and obtain the corresponding distribution probability of various disasters by performing operations at each level of the initial disaster distribution probability prediction model; The distribution probability of each type of disaster obtained and the label of the input target sample data are calculated through cross entropy loss to obtain the corresponding loss value; Update the weight parameters of the initial disaster distribution probability prediction model through back propagation; When the number of iterative trainings of the initial disaster distribution probability prediction model through the training set meets the first condition, the initial disaster distribution probability prediction model is tested and evaluated through the test set to obtain a corresponding evaluation result; When the evaluation result indicates that the initial disaster distribution probability prediction model is qualified, obtaining a corresponding disaster distribution probability prediction model; When the evaluation result indicates that the initial disaster distribution probability prediction model training is unqualified, a new round of training is performed on the initial disaster distribution probability prediction model using the training set.

7. The method for calculating the distribution probability of a composite flood disaster according to claim 5, characterized in that: Before inputting the value of the target disaster factor of the area to be predicted into the disaster distribution probability prediction model for processing and predicting the distribution probability of various disasters belonging to indirectly related disasters in the area to be predicted, the method further includes: Based on the correlation values ​​of the target disaster factors, all target disaster factors are sorted from large to small to obtain the sorting results; According to the ranking result, the target disaster factors with the smallest correlation value are removed from all target disaster factors in turn to form a new target disaster factor combination; Determine the AUC curve value corresponding to each new target disaster factor combination through the disaster distribution probability prediction model; The target disaster factor combination corresponding to the maximum AUC curve value is determined as the optimal disaster factor combination; The method of inputting the value of the target disaster factor of the area to be predicted into the disaster distribution probability prediction model for processing, and predicting the distribution probability of various disasters belonging to indirectly related disasters in the area to be predicted, includes: The values ​​of each disaster factor in the optimal disaster factor combination of the area to be predicted are input into the disaster distribution probability prediction model for processing, and the distribution probability of various disasters belonging to indirectly related disasters in the area to be predicted is predicted.

8. The method for calculating the distribution probability of a composite flood disaster according to claim 5, characterized in that: The chi-square test algorithm is used to determine the correlation values ​​between various disaster-causing factors and indirectly related disasters, including: By bringing the data of various disaster-causing factors into the chi-square test algorithm for calculation, the correlation values ​​of various disaster-causing factors and indirect related disasters are obtained. The expression of the chi-square test algorithm is: Among them, A is the number of hazard factors observed in various types of indirect related disasters, p refers to the expected probability under the assumption that the hazard factor has no significant difference in various types of indirect related disasters, k is the sum of the number of categories of the hazard factor corresponding to various types of indirect related disasters, It represents the correlation value of the disaster factor, E is the expected value under the assumption that the disaster factor has no significant difference in various types of indirect related disasters, and n is the frequency of occurrence of the disaster factor category.

9. The method for calculating the distribution probability of a composite flood disaster according to claim 1, characterized in that: The method further comprises: The disaster-prone objects and various disasters belonging to directly related disasters in the area to be predicted are combined with the various disasters belonging to indirectly related disasters obtained by prediction to obtain the distribution probability of various upstream and downstream disasters in the area to be predicted.

10. A distribution probability calculation system for compound flood disasters, characterized in that: The system comprises: The first construction module is used to construct a composite disaster knowledge graph based on disaster events that occurred within a preset time period before and after the flood disaster in the study area, wherein the nodes of the composite disaster knowledge graph are disaster events and disaster-bearing bodies, and the relationship between the nodes is the start time and the end time; The second construction module is used to extract the relationship chain related to flood disasters in the composite disaster knowledge graph based on the flood corpus, and construct the corresponding composite flood disaster knowledge graph; An extraction module is used to extract upstream disaster events pointing to flood disasters and downstream disaster events pointing to flood disasters in the composite flood disaster knowledge graph through grammatical statements; A correlation attribute determination module is used to determine the correlation attributes of the upstream disaster event and the downstream disaster event according to the respective occurrence probabilities of the upstream disaster event and the downstream disaster event, and the correlation attributes include directly related disasters and indirectly related disasters; A sample data determination module is used to obtain the value of the target disaster factor in the study area, determine the value as sample data, determine the upstream disaster events and downstream disaster events with correlation attributes in the study area as labels to mark the sample data, and obtain target sample data; The model training and testing module is used to train and test the constructed initial disaster distribution probability prediction model through a preset number of target sample data to obtain a qualified disaster distribution probability prediction model; The distribution probability prediction module is used to input the value of the target disaster-causing factor of the area to be predicted into the disaster distribution probability prediction model for processing, and predict the distribution probability of various disasters belonging to indirectly related disasters in the area to be predicted.

Images