Flood disaster risk analysis method and device, program product and electronic equipment

Abstract

The present disclosure relates to the technical field of risk analysis, and provides a flood disaster risk analysis method and device, a program product and an electronic device. The method comprises: preprocessing flood disaster record data to obtain a transaction dataset; determining an initial minimum support threshold according to the distribution of the support of a first item set in the transaction dataset; calculating adaptive minimum support thresholds of second item sets of different lengths based on the initial minimum support threshold to obtain a target frequent item set; splitting the target frequent item set into antecedent items and consequent items to generate a first candidate association rule set, and determining an initial minimum confidence threshold; calculating adaptive minimum confidence thresholds of different antecedent item lengths according to the initial minimum confidence threshold and the length of the antecedent item, and obtaining a target association rule according to the adaptive minimum threshold to perform flood disaster risk analysis based on the target association rule. The present scheme can improve the accuracy of flood disaster cause analysis.

Description

Technical Field

[0001] This disclosure relates to the field of risk analysis technology, and more specifically, to a flood disaster risk analysis method, a flood disaster risk analysis device, a computer program product, and an electronic device. Background Technology

[0002] Among related technologies, the analysis of the causes of watershed flood risk can be achieved by combining multi-source heterogeneous data fusion preprocessing with basic data mining algorithms. For example, data cleaning, feature extraction and transformation techniques can be used to integrate multi-source heterogeneous data such as rainfall, underlying surface, meteorology, and engineering status. Then, association rule mining algorithms can be used to mine the potential correlations between various disaster-causing factors. At the same time, a multi-level information fusion framework can be relied upon to reduce data loss and improve data availability.

[0003] While this approach can effectively expand the scope of data utilization and preliminarily uncover the coupling effects of various factors, thus improving the targeting of causal analysis, it lacks the ability to uncover weakly correlated causes and cannot accurately capture the potential correlation between niche disaster-causing factors and flood risk, resulting in an incomplete correlation mining.

[0004] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this disclosure is to provide a flood disaster risk analysis method and apparatus, computer program product and electronic device, thereby improving the accuracy of flood risk cause analysis to at least a certain extent.

[0006] Other features and advantages of this disclosure will become apparent from the following detailed description, or may be learned in part from practice of this disclosure.

[0007] According to a first aspect of this disclosure, a flood disaster risk analysis method is provided, comprising: acquiring flood disaster record data of different watershed types; preprocessing the flood disaster record data to obtain a transaction dataset corresponding to each watershed type, wherein the flood disaster record data includes at least one flood impact factor data and flood disaster severity data; for each watershed type, calculating the support of a first itemset of a first preset length in the transaction dataset of the watershed type, determining an initial minimum support threshold corresponding to different watershed types based on the distribution of the support of the first itemset, wherein the support of the itemset is the ratio of the number of transactions containing the itemset to the total number of transactions in the transaction dataset; for second itemsets of different lengths in each watershed type, calculating an adaptive minimum support threshold corresponding to the second itemset based on the initial minimum support threshold and the length of the second itemset, and filtering target frequent itemsets corresponding to each watershed type based on the adaptive minimum support threshold, wherein the itemsets... The length is the number of items contained in the itemset. For each watershed type, the target frequent itemset is split into antecedent and consequent items to generate a first candidate association rule set. Based on the distribution of confidence scores of candidate association rules of a second preset length in the first candidate association rule set, an initial minimum confidence threshold is determined. The confidence score is the ratio of the number of transactions containing both antecedent and consequent items to the number of transactions containing antecedent items. For each watershed type, the adaptive minimum confidence threshold of the second candidate association rule set is calculated based on the initial minimum confidence threshold and the antecedent item length. For each watershed type, the second candidate association rules in the second candidate association rule set are filtered based on the adaptive minimum confidence threshold of each second candidate association rule set to obtain the target association rules corresponding to each watershed type, so as to conduct flood disaster risk analysis based on the target association rules.

[0008] According to a second aspect of this disclosure, a flood disaster risk analysis device is provided, comprising: a preprocessing module configured to acquire flood disaster record data of different watershed types, preprocess the flood disaster record data to obtain a transaction dataset corresponding to each watershed type, wherein the flood disaster record data includes at least one flood impact factor data and flood disaster severity data; an initial minimum support threshold determination module configured to, for each watershed type, calculate the support of a first itemset of a first preset length in the transaction dataset of the watershed type, and determine an initial minimum support threshold corresponding to different watershed types based on the distribution of the support of the first itemset, wherein the support of the itemset is the ratio of the number of transactions containing the itemset to the total number of transactions in the transaction dataset; and a target frequent itemset determination module configured to... For each watershed type, a second itemset of different lengths is used. An adaptive minimum support threshold is calculated based on the initial minimum support threshold and the length of the second itemset. The adaptive minimum support threshold is then used to filter and obtain target frequent itemsets for each watershed type. The itemset length is the number of items contained in the itemset. An initial minimum confidence threshold determination module is configured to split the target frequent itemsets for each watershed type into antecedent and consequent items, generating a first candidate association rule set. Based on the confidence distribution of candidate association rules of a second preset length in the first candidate association rule set, an initial minimum confidence threshold is determined. The confidence level is the ratio of the number of transactions containing both antecedent and consequent items to the number of transactions containing only the antecedent item. The adaptive minimum confidence threshold determination module is configured to calculate the adaptive minimum confidence threshold of the second candidate association rule set based on the initial minimum confidence threshold and the antecedent length for each watershed type and the second candidate association rule set with different antecedent lengths. The flood disaster analysis module is configured to filter the second candidate association rules in the second candidate association rule set for each watershed type based on the adaptive minimum confidence threshold of each second candidate association rule set to obtain the target association rule corresponding to each watershed type, so as to perform flood disaster risk analysis based on the target association rule.

[0009] According to a third aspect of this disclosure, a computer program product comprising instructions is provided that, when run on a computer, causes the computer to perform the steps of the flood disaster risk analysis method as described in the first aspect.

[0010] According to a fourth aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the flood disaster risk analysis method as described in the first aspect of the above embodiments.

[0011] According to a fifth aspect of this disclosure, an electronic device is provided, comprising: a processor; and a storage device for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement the flood disaster risk analysis method as described in the first aspect of the above embodiments.

[0012] As can be seen from the above technical solutions, the flood disaster risk analysis method, the flood disaster risk analysis device, and the computer program product and electronic device for implementing the flood disaster risk analysis method in the exemplary embodiments of this disclosure have at least the following advantages and positive effects: In some embodiments of the present disclosure, the technical solutions provided include, on the one hand, calculating an adaptive minimum support threshold for itemsets of different lengths based on the distribution of support and the length of itemsets, and calculating an adaptive minimum confidence threshold for rules of different lengths based on the distribution of confidence and the length of the rule's antecedents. This avoids the problem of weak association rules being difficult to discover due to fixed thresholds, and improves the comprehensiveness of association rule mining. On the other hand, the present disclosure determines the minimum support threshold and minimum confidence threshold for each watershed type based on their respective data distribution, so that the mined target association rules can accurately reflect the flood disaster risk patterns of different watershed types. This avoids the omission of key rules or interference from redundant rules due to the mismatch between the threshold and the data characteristics of a specific watershed, improving the accuracy of association rule mining and thus improving the accuracy of flood risk causation analysis. Furthermore, since an adaptive threshold strategy is used in both the frequent itemset mining and association rule screening stages, the number of invalid candidate itemsets and redundant candidate rules is effectively reduced, computational overhead is reduced, and mining efficiency is improved.

[0013] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0014] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure. It is obvious that the drawings described below are merely some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0015] Figure 1 A flowchart illustrating a flood disaster risk analysis method according to an exemplary embodiment of this disclosure is shown. Figure 2 A flowchart illustrating a method for obtaining a target frequent itemset according to an exemplary embodiment of this disclosure is shown. Figure 3A flowchart illustrating a method for determining an adaptive minimum confidence threshold according to an exemplary embodiment of this disclosure is shown. Figure 4 A flowchart illustrating a method for determining target association rules according to an exemplary embodiment of this disclosure is shown. Figure 5 A flowchart illustrating a method for determining the basic impact weights of flood impact factors according to an exemplary embodiment of this disclosure is shown. Figure 6 This illustration shows a flowchart of a method for analyzing the causes of flood disaster risk according to an exemplary embodiment of this disclosure; Figure 7 This diagram illustrates the composition of a flood disaster risk analysis device according to an exemplary embodiment of the present disclosure; Figure 8 A schematic diagram of the structure of an electronic device according to an exemplary embodiment of the present disclosure is shown. Detailed Implementation

[0016] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this disclosure more comprehensive and complete, and to fully convey the concept of the example embodiments to those skilled in the art. The described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced with one or more of the specific details omitted, or other methods, components, apparatus, steps, etc., can be employed. In other instances, well-known technical solutions are not shown or described in detail to avoid obscuring various aspects of this disclosure.

[0017] The terms “a,” “an,” “the,” and “the” are used in this specification to indicate the presence of one or more elements / components / etc.; the terms “including” and “having” are used to indicate an open-ended inclusion and to mean that there may be other elements / components / etc. in addition to the listed elements / components / etc.; the terms “first” and “second” are used only as markings and are not a limitation on the number of objects.

[0018] Furthermore, the accompanying drawings are merely illustrative of this disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0019] In related technologies, the causes of flood disaster risk can be mined by combining multi-source heterogeneous data fusion preprocessing with basic data mining algorithms. For example, data cleaning, feature extraction, and transformation techniques can be used to integrate multi-source heterogeneous data such as rainfall, underlying surface, meteorological, and engineering conditions. Then, association rule mining algorithms can be used to uncover potential correlations between various disaster-causing factors. Simultaneously, a multi-level information fusion framework can be relied upon to reduce data loss and improve data usability. This type of technology can effectively expand the scope of data utilization and initially uncover the coupling effects of various factors, thus improving the targeting of causal analysis. However, it has the following drawbacks: the multi-source data fusion process is susceptible to data heterogeneity, leading to data mismatch and large errors. It also struggles to handle data noise, missing values, and other quality issues. Furthermore, the basic mining algorithms lack the ability to mine weakly correlated causes, failing to accurately capture the potential correlation between niche disaster-causing factors and flood risk, resulting in insufficiently comprehensive correlation mining.

[0020] Furthermore, related technologies employ data mining methods such as combined weighting and machine learning to quantify the weights of risk causes. For instance, game theory-based combined weighting integrates subjective and objective weights to quantify the impact of each disaster-causing factor on flood risk, while simple machine learning algorithms optimize prediction accuracy. While this approach overcomes the limitations of traditional qualitative analysis and simplifies the quantitative analysis process, making weight quantification more reasonable, it suffers from the following drawbacks: game theory-based combined weighting requires extremely high integrity of the basic data; when data is missing, the weight calculation deviation is significant; the algorithm has poor adaptability, making it difficult to adapt to the differences in disaster-causing characteristics across different watersheds; and it does not fully integrate real-time monitoring data for dynamic weight adjustment, resulting in insufficient real-time weight quantification and an inability to support accurate early warning and prevention.

[0021] To address one or more of the aforementioned problems, this disclosure provides a flood disaster risk analysis method, exemplarily, Figure 1 This illustration shows a flood disaster risk analysis method according to an exemplary embodiment of the present disclosure. (Reference) Figure 1 The method includes: Step S110: Obtain flood disaster record data for different watershed types, preprocess the flood disaster record data to obtain a transaction dataset corresponding to each watershed type, wherein the flood disaster record data includes at least one flood impact factor data and flood disaster severity data; Step S120: For each watershed type, the support of the first itemset of the first preset length in the transaction dataset of the watershed type is calculated, and the initial minimum support threshold corresponding to different watershed types is determined according to the distribution of the support of the first itemset. The support of the itemset is the ratio of the number of transactions containing the itemset to the total number of transactions in the transaction dataset. Step S130: For second itemsets of different lengths in each watershed type, calculate the adaptive minimum support threshold corresponding to the second itemset based on the initial minimum support threshold and the length of the second itemset, and filter the target frequent itemsets corresponding to each watershed type based on the adaptive minimum support threshold, wherein the itemset length is the number of items contained in the itemset; Step S140: For each type of watershed, the target frequent itemset is split into antecedent and consequent items to generate a first candidate association rule set. Based on the distribution of confidence of the candidate association rules of the second preset length in the first candidate association rule set, an initial minimum confidence threshold is determined. The confidence is the ratio of the number of transactions that contain both antecedent and consequent items to the number of transactions that contain antecedent items. Step S150: For the second candidate association rule set with different antecedent item lengths for each watershed type, calculate the adaptive minimum confidence threshold of the second candidate association rule set based on the initial minimum confidence threshold and the antecedent item length; Step S160: For each watershed type, the second candidate association rules in the second candidate association rule set are filtered according to the adaptive minimum confidence threshold of each second candidate association rule set to obtain the target association rule corresponding to each watershed type, so as to perform flood disaster risk analysis based on the target association rule.

[0022] exist Figure 1In the technical solution provided by the illustrated embodiment, on the one hand, this disclosure calculates an adaptive minimum support threshold for itemsets of different lengths based on the distribution of support and the length of itemsets, and calculates an adaptive minimum confidence threshold for rules of different lengths based on the distribution of confidence and the length of the rule's antecedent. This avoids the problem of weak association rules being difficult to discover due to fixed thresholds, and improves the comprehensiveness of association rule mining. On the other hand, this disclosure determines the minimum support threshold and minimum confidence threshold for each watershed type based on their respective data distribution, so that the mined target association rules can accurately reflect the flood disaster risk patterns of different watershed types. This avoids the omission of key rules or interference from redundant rules due to the mismatch between the threshold and the data characteristics of a specific watershed, improving the accuracy of association rule mining and thus improving the accuracy of flood risk causation analysis. Furthermore, since an adaptive threshold strategy is used in both the frequent itemset mining and association rule screening stages, the number of invalid candidate itemsets and redundant candidate rules is effectively reduced, computational overhead is reduced, and mining efficiency is improved.

[0023] Next, a detailed description will be given of the specific implementation method of "step S110, obtaining flood disaster record data of different watershed types, and preprocessing the flood disaster record data to obtain the transaction dataset corresponding to each watershed type".

[0024] In one exemplary embodiment, the flood disaster record data includes at least one flood impact factor data and flood disaster severity data. The flood impact factor data includes multiple data sources such as meteorological and hydrological data, geospatial data, human activity data, and water conservancy project status data.

[0025] Meteorological and hydrological data can include rainfall data. Multi-source heterogeneous data can be obtained through different methods, such as meteorological satellite remote sensing rainfall data, radar rainfall data, and ground rain gauge measured data. Rainfall data can include rainfall duration and rainfall amount.

[0026] Geospatial data can be understood as underlying surface data, which can include data such as soil type, topographic slope, and vegetation cover. Underlying surface data is spatial distribution data that characterizes the physical features of the watershed surface. It is mainly obtained through GIS (Geographic Information System) technology. For example, a GIS platform can be used to load multi-source spatial data such as Digital Elevation Model (DEM) data, soil type distribution maps, Land Use and Land Cover (LULC) maps, vegetation cover remote sensing products, and water system distribution maps of the target watershed. The spatial analysis functions of GIS (such as extraction analysis, reclassification, and raster calculation) can be used to process the above data and extract the underlying surface characteristic parameters corresponding to each sub-watershed or each monitoring station.

[0027] Human activity data refers to factor data characterizing the impact of human activities on flood formation and disaster risk, primarily obtained through statistical surveys combined with GIS spatialization. For example, statistical yearbooks, population census data, national economic statistics, and municipal infrastructure data from housing and urban-rural development departments of various administrative regions within a river basin can be collected. These statistical data can then be distributed to sub-basins or grid units using GIS spatial correlation and spatial interpolation methods (such as inverse distance weighted interpolation and Kriging interpolation), extracting the corresponding human activity characteristic parameters for each sub-basin. Human activity data can include one or more of the following: population asset distribution data (such as population density and fixed asset distribution), land use and cover change data (such as urbanization rate, cultivated land reclamation rate, and forest cover change rate), and soil and water conservation measures data (such as the distribution and quantity of terraces).

[0028] Hydraulic engineering status data refers to monitoring data characterizing the real-time operational status and structural safety of various hydraulic engineering projects within a watershed. This data is primarily acquired through automated monitoring systems for hydraulic engineering projects. For example, sensors such as piezometers, displacement gauges, strain gauges, water level gauges, and flow monitoring instruments deployed at hydraulic engineering sites can collect operational parameters in real time. These parameters are then aggregated into a data center via an automated monitoring system and verified and supplemented by manual inspection records. Hydraulic engineering data can include one or more of the following: reservoir operational parameters (such as reservoir water level, inflow, outflow, and storage capacity); dike monitoring data (such as water level at dike monitoring sections and seepage pressure within the dike); dam and gate operational parameters (including upstream and downstream water levels, flow through the gate, and gate opening / closing status); and flood storage and detention area status data (including flood storage and detention area activation status, water level within the area, and drainage flow).

[0029] Data on the severity of flood disasters can include historical flood data and disaster survey data recorded by various hydrological and water level stations within the basin, such as historical disaster level data, disaster-causing factor data, and disaster loss data (e.g., affected population, affected crop area, number of collapsed defenses, etc.).

[0030] For example, historical flood disaster records from different types of river basins can be obtained, such as flood disaster records from the last three years. Data cleaning techniques (missing value completion, outlier removal), feature standardization, and heterogeneous data adaptation and transformation techniques from data mining can be used to process the historical flood disaster records to solve problems such as data noise, missing values, and mismatches, generating a standardized dataset. For instance, when data is missing, the K-nearest neighbor difference algorithm can be used for completion. Of course, other methods can also be used for data completion; this exemplary implementation does not impose any special limitations on this.

[0031] For example, one implementation of step S110 may include: discretizing the flood disaster data for each watershed type, and generating a transaction dataset for that watershed type based on the discrete values ​​of the flood disaster record data.

[0032] For example, the discretized flood impact factor category labels and flood disaster severity level labels of each flood disaster record can be combined into a transaction record. In other words, each transaction record consists of multiple items, each item being a tuple of "factor name = factor value," such as "rainfall = heavy" being one item. The set of all items constitutes the transaction corresponding to that flood disaster record.

[0033] Taking the flood disaster record of a mountainous river as an example, the corresponding transaction record generated after discretization is: {Rainfall = H, Maximum 24-hour rainfall = H, Maximum 1-hour rainfall = M, Previous impact rainfall = H, River gradient = H, Watershed area = M, Vegetation coverage = L, Soil type = Loam, Peak flow modulus = H, Disaster level = Severe}, where H represents high level, L represents low level, and M represents medium level. That is, each transaction record is a set of different items, meaning each transaction record exists in the form of an itemset. The dataset composed of all transaction records is the transaction dataset. In other words, after converting all flood disaster record data into transaction record form, we can obtain a transaction dataset where each transaction exists in the form of an itemset, which can be identified by the association mining algorithm.

[0034] In one exemplary embodiment, different watershed types may include mountainous watersheds, plain watersheds, and small watersheds (remote watersheds). Of course, other watershed types may also be included, and this exemplary embodiment does not impose any special limitations on them.

[0035] Next, a detailed explanation will be given of the specific implementation method of "step S120, for each type of watershed, statistically analyze the support of the first itemset of the first preset length in the transaction dataset of the watershed type, and determine the initial minimum support threshold corresponding to different watershed types based on the distribution of the support of the first itemset".

[0036] In one exemplary implementation, an item is a discrete value of the flood impact factor at the smallest granularity. For example, "rainfall intensity = high" and "river level = high" are both single items. An itemset is a set formed by combining at least one item. For example, {rainfall intensity = heavy rain} constitutes a single itemset, and {rainfall intensity = heavy rain, river level = high} constitutes a binary itemset. The length of an itemset is the number of items it contains. For example, each transaction in the transaction dataset stores all items corresponding to a single flood event in the basin. A single transaction may contain 5 items, namely the values ​​corresponding to rainfall intensity, river level, soil moisture content, topographic slope, and water conservancy project status. An itemset containing only 1 item has a length of 1, an itemset containing 2 items has a length of 2, and so on.

[0037] In one exemplary implementation, the support of an itemset is the ratio of the number of transactions containing the itemset to the total number of transactions in the transaction dataset. For example, if an itemset is "Rainfall Intensity = High", and 30 transaction records in the transaction dataset contain this itemset, and the transaction dataset contains a total of 100 transaction records, then the support of this itemset is 0.3. Support is used to characterize how frequently this set of flood impact factor data combinations occurs in historical flood events.

[0038] In one exemplary implementation, the first preset length can be 1, that is, for each type of watershed transaction dataset, the support of itemsets of length 1 in the corresponding transaction dataset can be calculated separately.

[0039] A one-itemset contains only a single flood impact factor and does not depend on the simultaneous occurrence of other flood impact factors. Therefore, compared to itemsets of other lengths, one-itemsets have the highest frequency of occurrence and the largest transaction coverage. The statistically obtained support distribution is not affected by combinatorial sparsity. For example, if the distribution of three-itemsets is used to determine the initial minimum support threshold, the support itself is generally low because the combination of three flood impact factors may be extremely rare, and it cannot represent the overall factor occurrence pattern of the watershed. Therefore, the first preset length is set to 1, which allows the initial minimum support threshold to be determined based on the support distribution of all one-itemsets.

[0040] For example, the support of all one-items in the transaction dataset can be calculated, the distribution of the support of all one-items can be analyzed (such as quantiles, mean, median, distribution interval), and the initial minimum support threshold can be determined based on the distribution characteristics.

[0041] For example, first calculate the support distribution of all 1-itemsets, sort the support from high to low, find the inflection point of the support decline curve using the elbow rule, and use the support corresponding to the inflection point as the initial minimum support threshold; alternatively, the quantile method can be used, taking the 20th percentile of the support of all individual items as the initial minimum support threshold to ensure that most statistically significant items are covered.

[0042] In one exemplary implementation, the minimum support threshold is used to select frequent itemsets from the entire itemset of the transaction dataset whose support is greater than or equal to the minimum support threshold.

[0043] Next, a detailed description will be given of the specific implementation of "step S130, for each type of watershed, for a second itemset of different lengths, calculate the adaptive minimum support threshold corresponding to the second itemset based on the initial minimum support threshold and the length of the second itemset, and filter the target frequent itemsets corresponding to each type of watershed based on the adaptive minimum support threshold".

[0044] For each watershed type, it can be based on Figure 2 The methods shown yield their corresponding target frequent itemsets. For example, Figure 2 This diagram illustrates a flowchart of a method for obtaining a target frequent itemset according to an exemplary embodiment of this disclosure, with reference to... Figure 2 The method may include steps S210 to S220. Wherein: In step S210, for second itemsets of different lengths, an adaptive minimum support threshold corresponding to the second itemset is determined based on the ratio of the initial minimum support threshold to the length of the second itemset.

[0045] For example, the minimum support threshold for second itemsets of different lengths can be determined using the following formula (1): (1) In formula (1), The length of the second itemset, which can include all itemsets in the transaction dataset.

[0046] As can be seen from the above formula (1), the larger the itemset length, the smaller its corresponding adaptive minimum support threshold. Long itemsets correspond to disastrous combinations with multi-factor coupling, and their natural co-occurrence probability is much lower than that of single itemsets. Fixing a uniform threshold would cause most long itemsets with multi-factor coupling to be directly filtered out. Therefore, gradually reducing the minimum support threshold as the itemset length increases can ensure the statistical reliability of short itemsets while effectively recalling rare but highly disastrous multi-factor combinations.

[0047] Of course, other methods can also be used to determine the adaptive minimum support threshold corresponding to itemsets of different lengths based on the initial minimum support threshold and the itemset length. It is only necessary to satisfy that the larger the itemset length, the smaller the adaptive minimum support threshold. This exemplary implementation does not impose any special limitations on this.

[0048] In step S220, frequent itemsets with support greater than the adaptive minimum support threshold corresponding to the second itemset are selected from the second itemset to obtain the target frequent itemset.

[0049] For example, the support of each itemset in the transaction dataset can be calculated. For each itemset, an adaptive minimum support threshold can be determined based on its length. If the length of the itemset is greater than the adaptive minimum support threshold, then the itemset is a frequent itemset; otherwise, it is not. All frequent itemsets are then considered as target frequent itemsets.

[0050] Through steps S210 to S220 above, an adaptive mechanism is used to make the adaptive minimum support threshold lower as the itemset length increases. While retaining the weakly coupled disaster-causing pattern of long itemsets, a relatively higher screening threshold is maintained for short itemsets, eliminating low-value redundant itemsets. This balances the completeness of flood multi-factor coupling cause mining with the computational efficiency of subsequent rule mining, thereby improving the effectiveness of subsequent flood-causing association rule mining.

[0051] Next, a detailed explanation will be given of the specific implementation method of "step S140, for each type of watershed corresponding target frequent itemset, splitting the target frequent itemset into antecedent and consequent items to generate a first candidate association rule set, and determining the initial minimum confidence threshold according to the distribution of the confidence of the candidate association rules of the second preset length in the first candidate association rule set".

[0052] In one exemplary implementation, the confidence level is the ratio of the number of transactions that simultaneously include both the antecedent and the consequent to the number of transactions that include only the antecedent. The confidence level characterizes the disaster-causing reliability of the association rule, reflecting the probability that the corresponding flood disaster severity data will occur simultaneously, given the existence of flood impact factor data.

[0053] For example, a first-choice association rule might be "Rainfall = Heavy, Flood Disaster Severity = High". For the transaction dataset corresponding to the current watershed type, the total number of transactions containing the antecedent "Rainfall = Heavy" is 50. Among these, 35 transactions contain both the antecedent "Rainfall = Heavy" and the consequent "Flood Disaster Severity = High". According to the confidence calculation rule, the confidence of this candidate association rule = (Number of transactions containing both antecedent and consequent) / (Number of transactions containing only the antecedent) = 35 / 50 = 0.7. This result indicates that in historical flood events in this watershed, there is a 70% probability of a high-level flood disaster occurring under heavy rainfall conditions. A higher confidence score indicates a stronger correlation between the factors and the corresponding flood disaster, and a higher rule confidence.

[0054] For example, one can first remove the target frequent itemsets that do not contain flood risk disaster level data from the target frequent itemsets, and then use the flood impact factor data in the remaining target frequent itemsets as antecedents and the flood disaster level data as consequents, thereby obtaining the first candidate association rules corresponding to the remaining target frequent itemsets. The first candidate association rule set is obtained based on the set of all the first candidate association rules.

[0055] In one exemplary implementation, the second preset length is 2, meaning the initial minimum support threshold can be obtained based on the distribution of confidence scores of all single-factor disaster-causing association rules in the first association rule set. The first candidate association rule of length 2 can be decomposed into the simplest causal association rule corresponding to a single flood disaster severity for a single flood impact factor, representing the minimum structure capable of generating effective disaster-causing rules. This type of rule has a sufficient number of samples and a stable confidence distribution, objectively reflecting the basic disaster-causing probability of the watershed. Therefore, it can be used to determine the initial minimum confidence threshold. Longer multi-factor rules, on the other hand, have sparse samples and strong randomness, making them unsuitable as a confidence threshold benchmark.

[0056] For example, first, statistically analyze the confidence distribution of all candidate association rules causing single-factor disasters, sort the confidence levels from high to low, and use the elbow rule to find the inflection point of the support decline curve. Use the confidence level corresponding to the inflection point as the initial minimum confidence threshold. Alternatively, the quantile method can be used, taking the 20th percentile of all candidate association rules causing single-factor disasters as the initial minimum confidence threshold. In this way, while filtering out low-confidence noise rules, most of the statistically significant basic disaster-causing rules are retained, providing a stable baseline for the subsequent adaptive confidence correction of multiple antecedent length rules.

[0057] Next, a detailed description will be given of the specific implementation of "step S150, for the second candidate association rule set with different antecedent item lengths for each watershed type, calculate the adaptive minimum confidence threshold of the second candidate association rule set according to the initial minimum confidence threshold and the antecedent item length".

[0058] For example, Figure 3 This diagram illustrates a flowchart of a method for determining an adaptive minimum confidence threshold according to an exemplary embodiment of this disclosure. (See also:) Figure 3 The method may include steps S310 and S330. Specifically: in step S310, the difference between the length of the antecedent item and the second preset length is calculated; in step S320, a first product of the difference and a preset coefficient is calculated; and in step S330, an adaptive minimum confidence threshold for the second candidate association rule set is obtained based on the sum of the first product and the initial minimum confidence threshold.

[0059] For example, the adaptive minimum confidence threshold corresponding to the second candidate association rule set can be obtained by the following formula (2): (2) In formula (2), 2 represents the length of the preceding item, and 2 represents the second preset length mentioned above. Preset coefficients, preset coefficients The value can be customized according to requirements, such as 0.03-0.08. The preset coefficient is used to control the increase in the minimum confidence threshold for different antecedent length rules.

[0060] Due to the length of the preceding piece The larger the antecedent length, the fewer the number of transaction samples that simultaneously satisfy all antecedent factors. This can lead to artificially high confidence levels based on only a small number of accidental co-occurrence samples, resulting in rules that lack general disaster-causing reference value. Therefore, a preset coefficient is introduced to make the confidence threshold increase linearly with the antecedent length, thereby filtering out overfitting pseudo-rules caused by small samples while mining multi-factor coupled disaster-causing rules.

[0061] For example, α = 0.05, when When = 2, the adaptive minimum confidence threshold is equal to the initial minimum confidence threshold. When the minimum confidence threshold is 3, the minimum confidence threshold is increased by 0.05 based on the initial minimum confidence threshold; when When the coefficient is 4, the minimum confidence threshold is increased by 0.10 based on the initial minimum confidence threshold, and so on, gradually increasing the minimum confidence threshold. If there are few historical disaster samples in the watershed, a smaller coefficient of 0.03 to 0.04 can be selected to reduce the stringency of the screening; if the watershed data volume is large and the factor dimensions are complex, a larger coefficient of 0.06 to 0.08 can be selected to enhance the pseudo-rule filtering effect.

[0062] Next, a detailed explanation will be given of the specific implementation method of "step S160, for each watershed type, according to the adaptive minimum confidence threshold of each second candidate association rule set, the second candidate association rule in the second candidate association rule set is filtered to obtain the target association rule corresponding to each watershed type, so as to perform flood disaster risk analysis based on the target association rule".

[0063] For example, for each type of watershed, it can be done through... Figure 4 The method shown yields the target association rules corresponding to this watershed type.

[0064] For example, Figure 4 This diagram illustrates a flowchart of a method for determining target association rules according to an exemplary embodiment of this disclosure. (See reference...) Figure 4 The method may include steps S410 to S430, wherein: In step S410, a third candidate association rule with a confidence level greater than the adaptive minimum confidence threshold of the second candidate association rule set is selected from each second candidate association rule set to obtain a third candidate association rule set.

[0065] For example, the first candidate association rules in the first candidate association rule set can be grouped according to the length of the antecedent, grouping first candidate association rules with the same antecedent length into the same group, thus obtaining second candidate association rule sets corresponding to different antecedent lengths. For example, an antecedent length of 1 corresponds to one second candidate association rule set, an antecedent length of 2 corresponds to another second candidate association rule set, and so on. For each second candidate association rule set, third candidate association rules with a confidence level greater than the corresponding adaptive minimum confidence threshold can be selected from the second candidate association rule set based on the adaptive minimum confidence threshold corresponding to the antecedent length of the second candidate association rules in that second candidate association rule set. The set composed of all third candidate association rules is then used to obtain the third candidate association rule set.

[0066] In step S420, for each third candidate association rule in the third candidate association rule set, the frequency ratio of the third candidate association rule is calculated.

[0067] In one exemplary implementation, the frequency ratio is the ratio of the support of the third candidate association rule to the product of the support of the antecedent and the support of the consequent. For example, the frequency ratio of each third candidate association rule can be obtained by the following formula (3): (3) In formula (3), sup3 is the support of the third candidate association rule, sup1 is the support of the antecedent in the third candidate association rule, and sup2 is the support of the consequent in the third candidate association rule.

[0068] In one exemplary implementation, the frequency ratio can be understood as the ratio of the probability of occurrence of a flood disaster level corresponding to a rule consequent under the condition of the rule antecedent to the global base probability of occurrence of that flood disaster level in all transactions.

[0069] In step S430, the third candidate association rule whose frequency ratio is greater than or equal to a preset frequency threshold is determined as the target association rule.

[0070] In one exemplary embodiment, the preset frequency ratio threshold can be customized as needed, and this exemplary embodiment does not impose any special limitations on it.

[0071] Relying solely on support and confidence levels can only screen for factor combinations that are statistically frequent and highly reliable, but it cannot distinguish whether the factors have a disaster-promoting effect or simply have a natural symbiotic relationship. If the probability of flood disasters is high within the watershed, some purely symbiotic factor combinations can also obtain high confidence levels, easily generating pseudo-rules with no practical value. Frequency ratio can quantify the extent to which a factor combination increases the probability of flood disasters. A frequency ratio greater than 1 indicates that the factor combination increases the probability of floods and has disaster warning significance; a frequency ratio close to 1 indicates that the factor combination has no obvious causal relationship with flood occurrence and is merely a natural co-occurrence, and should be eliminated.

[0072] In this disclosure, by using frequency ratio, co-existing noise rules without causal relationship can be further filtered on the basis of adaptive support and adaptive confidence two-layer screening, and finally target association rules with real disaster-causing mechanism and can be used for watershed risk assessment are retained, thereby improving the reliability and accuracy of association rule mining.

[0073] For example, Figure 5 This diagram illustrates a flowchart of a method for determining the basic impact weights of flood impact factors according to an exemplary embodiment of this disclosure. (See also:) Figure 5 The method may include steps S510 to S550. Wherein: In step S510, the influence weight pairs corresponding to each pair of flood influence factors in the flood influence factors determined by experts are obtained.

[0074] For example, domain experts can perform pairwise comparisons of flood impact factors to obtain the corresponding impact weight pairs for each pair of flood impact factors. For instance, by comparing all flood impact factors pairwise, experts can analyze the degree of influence of each flood impact factor on flood disasters, thereby obtaining different impact weight pairs. Through pairwise comparisons, the impact weights of different flood impact factors can be distinguished more accurately.

[0075] In step S520, an influence weight matrix corresponding to the flood influence factor is constructed based on the influence weight pairs.

[0076] For example, based on the pairwise influence weights of all flood impact factors obtained in step S510, a square matrix is ​​constructed by arranging them in the order of the flood impact factors (the order can be customized according to needs). This is the influence weight matrix. In the matrix, the element in the i-th row and j-th column represents the importance score of the i-th factor relative to the j-th factor. The diagonal elements are all 1 (a factor is equally important to itself), and the symmetrical elements are reciprocals of each other. For example, if there are 5 flood impact factors, a 5th-order square matrix can be constructed.

[0077] In step S530, for each row of the influence weight matrix, the arithmetic square root of the product of the row data corresponding to the row is calculated, and the arithmetic square root is normalized to obtain the subjective influence weight of the flood influence factor represented by the row.

[0078] For example, the subjective influence weights of each flood influencing factor can be solved using the AHP (Analytic Hierarchy Process) square root method. First, for each row of the influence weight matrix, the arithmetic square root of the product of all matrix elements in that row is calculated; then, the arithmetic square roots corresponding to all factors are globally normalized to finally obtain the subjective influence weights of each flood influencing factor.

[0079] In step S540, a target decision model is trained based on flood disaster record data, and the objective impact weight of flood impact factors is determined based on the global feature dimension weights output by the target decision model.

[0080] For example, for each type of watershed, flood disaster records for that type can be obtained. Data on each flood influencing factor can be used as input feature data, and flood disaster severity data as labels, to train a target decision model for feature importance learning. After model training, the global feature dimension weights corresponding to each input feature are output; these weights represent the objective influence weights of each flood influencing factor. These weights are entirely derived from the distribution of real data and statistical analysis of real disaster contribution, eliminating subjective biases from human experience and accurately reflecting the actual influence of each factor in historical data.

[0081] In step S550, the basic influence weight of the flood influence factor is determined based on the subjective influence weight and the objective influence weight.

[0082] In one exemplary implementation, the subjective and objective impact weights of the same flood impact factor are weighted and fused together to obtain the final basic impact weight of each flood impact factor by combining experience from the human domain with objective laws from historical data. The weighting coefficients can be customized based on experience or requirements; this exemplary implementation does not impose any special limitations on this.

[0083] For example, Figure 6 This diagram illustrates a flowchart of a method for analyzing the causes of flood disaster risk according to an exemplary embodiment of this disclosure. (See reference...) Figure 6 The method may include steps S610 to S660, wherein: In step S610, the target association rule corresponding to the watershed type of the target watershed and the flood impact factor monitoring data of the target watershed within the most recent preset time period are obtained.

[0084] For example, the target watershed can be a watershed requiring flood disaster risk causation analysis, which can be customized according to needs. Through steps S110 to S160 above, target association rules corresponding to different watershed types are generated. For a target watershed, the target association rules corresponding to its watershed type can be combined with basic impact weights and real-time monitoring data to conduct flood risk causation analysis of the target watershed.

[0085] For example, the most recent preset duration can be data from the most recent 1 hour or 2 hours, which can be customized according to needs. This exemplary embodiment does not impose any special limitations on this.

[0086] In step S620, the flood impact factor monitoring data and the antecedents in the target association rule corresponding to the watershed type of the target watershed are matched to determine the target flood impact factor data indicated by the successfully matched antecedents.

[0087] For example, real-time flood impact factor monitoring data can be precisely matched against the antecedents of all target association rules corresponding to the target watershed. By iterating through all factor combinations and discrete threshold conditions of rule antecedents, rule antecedents that can be completely matched by the current real-time monitoring data are selected. The factor data corresponding to the successfully matched antecedent is the target flood impact factor data that is effective in the current time period.

[0088] In step S630, based on the flood impact factor monitoring data, the information entropy of the target flood impact factor indicated by the target flood impact factor data within the most recent preset time period is calculated.

[0089] Information entropy is used to quantify the degree of dispersion of a single factor's numerical fluctuations within a recent time period. The greater the fluctuation of a factor's numerical value and the more disordered its temporal distribution, the higher the information entropy value, indicating that the factor is sensitive to recent changes in its state and has a stronger dynamic influence on the current flood risk. Conversely, the lower the entropy value, the more stable the factor's state and the weaker its dynamic influence. In other words, the information entropy can be obtained based on the dispersion of the target flood impact factor within the recent preset time period; the greater the dispersion, the greater the information entropy.

[0090] In step S640, the basic influence weight of the target flood influence factor is adjusted according to the information entropy to obtain the real-time influence weight of the target flood influence factor.

[0091] In one exemplary implementation, using the basic influence weights of each factor obtained in step S550 as a fixed benchmark, and combining them with the time-series information entropy calculated in step S630, the weights of the currently effective target flood influence factors are dynamically adjusted. This weakens the weight proportion of stable and unchanging factors and increases the weight proportion of fluctuation-sensitive factors, ultimately adapting to the current basin conditions to obtain the real-time influence weights of each factor. In other words, the basic influence weights of target influence factors with high information entropy can be increased, while the basic influence weights of target influence factors with low information entropy can be decreased, thereby obtaining the real-time influence weights of each target influence factor. Factors that have not yet been matched and become effective do not participate in this weight adjustment calculation and do not affect the current risk assessment results.

[0092] In step S650, a visual analysis report is generated based on the target flood impact factor data and the corresponding real-time impact weights of the target flood impact factors.

[0093] In one exemplary implementation, a watershed flood risk visualization analysis report can be automatically generated based on the target flood impact factor data determined in step S620, the real-time impact weights of each factor obtained in step S640, and the factor time-series change curves within a preset time period, as well as comparative data of similar historical disaster-causing scenarios. The report can display the current core disaster-causing factors, the real-time impact ratio of each factor, the time-series evolution trend of factors, and the results of risk trigger analysis, intuitively presenting the formation mechanism and dynamic characteristics of the current watershed flood risk.

[0094] In step S660, the current flood disaster risk level of the target watershed is predicted based on the target flood disaster severity data in the target flood impact factor data.

[0095] In one exemplary implementation, the flood disaster severity data of the consequent item corresponding to the target association rule that was successfully matched in step S620 can be used as the predicted current flood disaster risk level of the target watershed.

[0096] For example, this disclosure improves the Apriori association rule mining algorithm by determining adaptive minimum support and adaptive minimum confidence thresholds. Based on this improved algorithm, deep mining is performed on the standardized dataset obtained after data cleaning and missing value completion. This uncovers potential associations and coupling mechanisms among various disaster-causing factors (rainfall intensity, soil type, vegetation cover, engineering operation status, etc.), generating an association rule set and clarifying the association between core and secondary disaster-causing factors. This addresses the problems of incomplete association mining and insufficient capture of weak associations in existing technologies. The Apriori association rule mining algorithm refers to mining association rules using the Apriori (prior) algorithm. Apriori is a classic algorithm in association rule mining; its core function is to efficiently find all strong association rules that satisfy minimum support and minimum confidence from a transaction database.

[0097] For example, in this disclosure, the AHP (Analytic Hierarchy Process) method and the RF (Random Forest) algorithm can be integrated to construct a dynamic weight quantification model. For instance, subjective weights are determined by combining the AHP algorithm with expert experience, while the RF algorithm is used to train on massive amounts of historical and real-time monitoring data to explore the impact of various factors on flood risk and determine objective weights. A dynamic correction coefficient is introduced (adjusted every 1-2 hours based on the update frequency of real-time monitoring data) to achieve dynamic weight quantification, solving the problems of inaccurate weight quantification, insufficient real-time performance, and poor adaptability in existing technologies.

[0098] For example, in this disclosure, the results of correlation analysis and dynamic weight quantification can be integrated to generate a visual analysis report, which clarifies the core disaster-causing factors, the influence weight of each factor, and the causal coupling mechanism. At the same time, a data interface is set up to synchronize the analysis results to the existing flood warning platform (adapted to multi-source data fusion, multi-channel transmission, and edge computing warning terminals), providing targeted support for the optimization of warning thresholds and the formulation of prevention and control measures, and solving the problems of the disconnect between existing technologies and the early warning system and the lack of practicality.

[0099] For example, in this disclosure, the analysis of the causes of flood disaster risk can be achieved through the following process. Specifically, firstly, multi-source heterogeneous data is collected (i.e., four-dimensional data including meteorological and hydrological data, underlying surface and topography, human activities and disaster-bearing bodies, engineering and prevention and control data, and historical disaster data); then, multi-source data fusion and preprocessing are performed, namely, multi-source data fusion, data cleaning, feature standardization (global / local), heterogeneous data transformation, and missing value completion (such as KNN (K-Nearest Neighbors) interpolation or GWR (Geographically Weighted Regression) interpolation to complete missing values), thereby generating a standardized dataset; then, association rule mining is performed based on an improved association rule algorithm, i.e., the standardized dataset is input into IM-Apriori (Improved Association Rule Mining). The Apriori algorithm (an improved version) performs multi-dimensional association mining through adaptive threshold adjustment, followed by frequency ratio verification and pruning optimization (e.g., if the number exceeds a preset number, target association rules that meet the preset number are selected based on confidence) to generate an association rule set. Then, dynamic weight quantification is performed: the AHP algorithm determines the subjective weights of the influencing factors in the association rules, the RF algorithm determines the objective weights, and the BOA (Bayesian Optimization Algorithm) dynamically adjusts the weighting coefficients of the subjective and objective weights to obtain the basic influence weights of the influencing factors. The basic influence weights of the influencing factors are adjusted using local entropy (as mentioned earlier, the information entropy within the most recent preset time period), outputting the dynamic real-time influence weight values ​​of the influencing factors. Finally, the results are integrated and output, generating a "cause-association-weight-spatial distribution" visual analysis report, and performing risk level classification and disaster situation prediction.

[0100] In this disclosure, the early warning system can also be adapted to the digital twin, that is, synchronized to the early warning platform / digital twin system through a standardized interface, thereby supporting the optimization of early warning thresholds and providing prevention and control decision suggestions.

[0101] In one exemplary embodiment of this disclosure, disaster cause analysis is performed by collecting watershed topographic data (30m resolution), meteorological monitoring data (2020-2024, sampling frequency 10 minutes / time), hydrological monitoring data (river water level and flow, sampling frequency 5 minutes / time), pipeline network data (32km of main pipeline and 87km of branch pipeline), and historical flood disaster data (2020-2024, a total of 17 flood events, including disaster level, impact range, and cause records). Five years of monitoring data (approximately 1.2 million records) were collected. Preprocessing took 28 minutes, including 11 minutes for data cleaning, 8 minutes for standardization and normalization, and 9 minutes for data dimensionality reduction. The data validity rate after preprocessing was ≥98%. 47 effective association rules were identified through association rule mining (e.g., "Short-term heavy rainfall ≥25mm / h + terrain slope ≤5° increases the risk of urban flooding by 83%"). K-means clustering was used to divide the data into 5 risk units (2 high-risk, 2 medium-risk, and 1 low-risk). Random forest algorithm was used to quantify the core causal weights (short-term heavy rainfall 38%, low-lying terrain 27%, pipeline congestion 19%, vegetation coverage 8%, and others 8%). Based on verification from 17 historical flood events, the core cause fit rate was 93.2%. After 72 hours of continuous operation, the system did not crash or lose data, the data transmission delay was ≤2 seconds, and the consistency of the analysis results was ≥99%. It can be seen that in this disclosure, the closed-loop process of "data preprocessing - multi-algorithm collaborative data mining - cause verification" solves the core pain points of traditional watershed flood disaster risk cause analysis, which are "subjective, inefficient, unquantifiable, and weak against interference", and improves the accuracy and efficiency of the analysis.

[0102] Furthermore, it should be noted that the above figures are merely illustrative representations of the processes included in the method according to exemplary embodiments of the present invention, and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Additionally, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0103] Furthermore, an exemplary embodiment of this disclosure also provides a flood disaster risk analysis device. (See reference...) Figure 7As shown, the flood disaster risk analysis device 700 includes the following program modules: a preprocessing module 710, configured to acquire flood disaster record data of different watershed types, preprocess the flood disaster record data to obtain a transaction dataset corresponding to each watershed type, wherein the flood disaster record data includes at least one flood impact factor data and flood disaster severity data; an initial minimum support threshold determination module 720, configured to, for each watershed type, calculate the support of a first itemset of a first preset length in the transaction dataset of the watershed type, and determine the initial minimum support threshold corresponding to different watershed types based on the distribution of the support of the first itemset, wherein the support of the itemset is the ratio of the number of transactions containing the itemset to the total number of transactions in the transaction dataset; and a target frequent itemset determination module 730, configured to, for each watershed type, calculate the adaptive minimum support threshold corresponding to the second itemset based on the initial minimum support threshold and the length of the second itemset, and filter the target frequent itemsets corresponding to each watershed type based on the adaptive minimum support threshold, wherein the itemset length is the length of the itemset package. The system includes: a minimum confidence threshold determination module 740, configured to split the target frequent itemset into antecedent and consequent items for each watershed type, generating a first candidate association rule set, and determining an initial minimum confidence threshold based on the confidence distribution of candidate association rules of a second preset length in the first candidate association rule set, wherein the confidence is the ratio of the number of transactions containing both antecedent and consequent items to the number of transactions containing antecedent items; an adaptive minimum confidence threshold determination module 750, configured to calculate an adaptive minimum confidence threshold for the second candidate association rule set with different antecedent item lengths for each watershed type, based on the initial minimum confidence threshold and the antecedent item length; and a flood disaster analysis module 760, configured to filter the second candidate association rules in the second candidate association rule set for each watershed type based on the adaptive minimum confidence threshold of each second candidate association rule set, obtaining the target association rule corresponding to each watershed type, and performing flood disaster risk analysis based on the target association rule.

[0104] The specific details of each part of the above-mentioned device have been described in detail in the method section of the implementation plan. For any undisclosed details, please refer to the implementation plan of the method section, and therefore will not be repeated here.

[0105] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to exemplary embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0106] Furthermore, although the steps of the method in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0107] An exemplary embodiment of this disclosure also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the flood disaster risk analysis method described above.

[0108] In one implementation, the computer program product can be a tangible product containing a computer program, such as a computer-readable storage medium storing the computer program. The readable storage medium can be a storage medium based on electrical, magnetic, optical, electromagnetic, infrared, or other signals, including but not limited to: random access memory (RAM), read-only memory (ROM), magnetic tape, floppy disk, flash memory, hard disk drive (HDD), solid-state drive (SSD), etc. For example, the computer program product can be implemented as a non-volatile storage medium storing a computer program, such as read-only memory, NAND flash memory, etc.

[0109] In one implementation, the computer program product can be an intangible product containing a computer program. For example, the computer program product can be implemented as a virtual digital product, such as an executable file, installation package, or other digital file storing the computer program.

[0110] Computer program code can be written in one or more programming languages. Examples of programming languages ​​include C, Java, C++, and Python. Program code can execute entirely on the user's computing device, partially on the user's computing device, or as a standalone software package. It can also execute partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, such as a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via an internet connection provided by a mobile network operator).

[0111] Computer programs can be carried or transmitted via signals such as electricity, magnetism, light, electromagnetic fields, and infrared radiation. Electronic devices can convert the signals carrying computer programs into digital signals, thereby running the computer programs. When a computer program runs on an electronic device, its code is used to cause the electronic device to execute (more specifically, to execute by the processor of the electronic device) the method steps of various exemplary embodiments of this disclosure, such as the flood disaster risk analysis method described above.

[0112] Exemplary embodiments of this disclosure also provide an electronic device, which may include a processor and a memory. The memory stores executable instructions of the processor, such as a computer program. The processor executes the executable instructions to perform the method steps of various exemplary embodiments of this disclosure. Furthermore, the electronic device may also include a display for displaying a graphical user interface.

[0113] The following is for reference. Figure 8 The electronic device is illustrated by way of a general-purpose computing device. It should be understood that... Figure 8 The electronic device 800 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0114] like Figure 8 As shown, the electronic device 800 may include: a processor 810, a memory 820, a bus 830, an I / O (input / output) interface 840, a network adapter 850, and a display 860.

[0115] The memory 820 may include volatile memory, such as RAM 821 and cache unit 822, and may also include non-volatile memory, such as ROM 823. The memory 820 may also include one or more program modules 824, including but not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. For example, program module 824 may include the modules described above.

[0116] The processor 810 may include one or more processing units, such as an AP (Application Processor), a modem processor, a GPU (Graphics Processing Unit), an ISP (Image Signal Processor), a controller, an encoder, a decoder, a DSP (Digital Signal Processor), a baseband processor, and / or an NPU (Neural-Network Processing Unit).

[0117] The processor 810 can be used to execute executable instructions stored in the memory 820, such as the flood disaster risk analysis method described above.

[0118] Bus 830 is used to connect different components of electronic device 800 and may include data bus, address bus and control bus.

[0119] Electronic device 800 can communicate with one or more external devices 900 (such as keyboard, mouse, external controller, etc.) through I / O interface 840.

[0120] Electronic device 800 can communicate with one or more networks via network adapter 850. For example, network adapter 850 can provide mobile communication solutions such as 3G / 4G / 5G, or wireless communication solutions such as wireless LAN, Bluetooth, and near-field communication. Network adapter 850 can communicate with other modules of electronic device 800 via bus 830.

[0121] Electronic device 800 can display a graphical user interface, such as a visual analysis report interface, through display 860.

[0122] although Figure 8As not shown in the diagram, other hardware and / or software modules may also be configured in the electronic device 800, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0123] Those skilled in the art will understand that various aspects of this disclosure can be implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which may be referred to as "circuit", "module" or "system" respectively.

[0124] It should be understood that this disclosure is not limited to the specific methods, steps, or structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. Those skilled in the art will readily conceive of other embodiments based on the specific implementations provided in this disclosure. Therefore, the specific implementations provided in this disclosure are merely exemplary, and the scope and spirit of this disclosure are indicated by the claims, and should cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary technical means in the art not disclosed in this disclosure.

Claims

1. A method for analyzing flood disaster risks, characterized in that, include: Acquire flood disaster record data for different watershed types, preprocess the flood disaster record data to obtain a transaction dataset corresponding to each watershed type, the flood disaster record data includes at least one flood impact factor data and flood disaster severity data; For each watershed type, the support of a first itemset of a first preset length in the transaction dataset of that watershed type is calculated. Based on the distribution of the support of the first itemset, the initial minimum support threshold corresponding to different watershed types is determined. The support of the itemset is the ratio of the number of transactions containing the itemset to the total number of transactions in the transaction dataset. For each watershed type, a second itemset of different lengths is used. An adaptive minimum support threshold is calculated based on the initial minimum support threshold and the length of the second itemset. The target frequent itemsets for each watershed type are then selected based on the adaptive minimum support threshold. The length of the itemset is the number of items contained in the itemset. For each type of watershed, the target frequent itemset is split into antecedent and consequent items to generate a first candidate association rule set. Based on the distribution of confidence of the candidate association rules of a second preset length in the first candidate association rule set, an initial minimum confidence threshold is determined. The confidence is the ratio of the number of transactions that contain both antecedent and consequent items to the number of transactions that contain antecedent items. For each watershed type with different antecedent item lengths, an adaptive minimum confidence threshold for the second candidate association rule set is calculated based on the initial minimum confidence threshold and the antecedent item length. For each watershed type, the second candidate association rules in the second candidate association rule set are filtered according to the adaptive minimum confidence threshold of each second candidate association rule set to obtain the target association rule corresponding to each watershed type, so as to conduct flood disaster risk analysis based on the target association rule.

2. The method according to claim 1, characterized in that, For each watershed type, the second itemset of different lengths is used to calculate an adaptive minimum support threshold corresponding to the second itemset based on the initial minimum support threshold and the length of the second itemset. The target frequent itemsets corresponding to each watershed type are then selected based on the adaptive minimum support threshold, including: For each watershed type, perform the following procedures to obtain the target frequent itemsets for each watershed type: For second itemsets of different lengths, an adaptive minimum support threshold corresponding to the second itemset is determined based on the ratio of the initial minimum support threshold to the length of the second itemset. From the second itemset, select frequent itemsets whose support is greater than the adaptive minimum support threshold corresponding to the second itemset to obtain the target frequent itemset.

3. The method according to claim 1, characterized in that, The adaptive minimum confidence threshold for the second candidate association rule set is calculated based on the initial minimum confidence threshold and the antecedent item length, including: Calculate the difference between the length of the preceding item and the second preset length; Calculate the first product of the difference and the preset coefficient; The adaptive minimum confidence threshold of the second candidate association rule set is obtained based on the sum of the first product and the initial minimum confidence threshold.

4. The method according to claim 1, characterized in that, For each watershed type, the second candidate association rules in the second candidate association rule set are filtered according to the adaptive minimum confidence threshold of each second candidate association rule set to obtain the target association rule corresponding to each watershed type, including: For each watershed type, the following process is performed separately to obtain the target association rules corresponding to each watershed type: From each set of second candidate association rules, select third candidate association rules whose confidence is greater than the adaptive minimum confidence threshold of the second candidate association rule set to obtain the third candidate association rule set; For each third candidate association rule in the third candidate association rule set, calculate the frequency ratio of the third candidate association rule, wherein the frequency ratio is the ratio of the support of the third candidate association rule to the product of the support of the antecedent item and the support of the consequent item; The third candidate association rule whose frequency ratio is greater than or equal to a preset frequency threshold is determined as the target association rule.

5. The method according to claim 1, characterized in that, The method further includes: Obtain the influence weight pairs corresponding to each pair of flood influence factors in the flood influence factors determined by experts; Construct the influence weight matrix corresponding to the flood influence factor based on the influence weight pairs; For each row in the influence weight matrix, the arithmetic square root of the product of the row data corresponding to the row is calculated, and the arithmetic square root is normalized to obtain the subjective influence weight of the flood influence factor represented by the row. A target decision model is trained based on flood disaster record data, and the objective impact weights of flood impact factors are determined based on the global feature dimension weights output by the target decision model. Based on the subjective influence weight and the objective influence weight, the basic influence weight of the flood influence factor is determined.

6. The method according to claim 5, characterized in that, The flood disaster risk analysis based on the target association rule includes: Obtain the target association rules corresponding to the watershed type of the target watershed and the flood impact factor monitoring data of the target watershed within the most recent preset time period; The flood impact factor monitoring data is matched with the antecedents in the target association rule corresponding to the watershed type of the target watershed to determine the target flood impact factor data indicated by the successfully matched antecedents; Based on the flood impact factor monitoring data, calculate the information entropy of the target flood impact factor indicated by the target flood impact factor data within the most recent preset time period; The basic influence weights of the target flood impact factors are adjusted based on the information entropy to obtain the real-time influence weights of the target flood impact factors. Based on the target flood impact factor data and the corresponding real-time impact weights of the target flood impact factors, a visual analysis report is generated; Predict the current flood disaster risk level of the target watershed based on the target flood disaster severity data in the target flood impact factor data.

7. The method according to claim 1, characterized in that, The flood impact factor data includes multiple data sources such as meteorological and hydrological data, geospatial data, human activity data, and water conservancy project status data.

8. A flood disaster risk analysis device, characterized in that, include: The preprocessing module is configured to acquire flood disaster record data of different watershed types, preprocess the flood disaster record data to obtain a transaction dataset corresponding to each watershed type, and the flood disaster record data includes at least one flood impact factor data and flood disaster severity data; The initial minimum support threshold determination module is configured to, for each watershed type, calculate the support of a first itemset of a first preset length in the transaction dataset of the watershed type, and determine the initial minimum support threshold corresponding to different watershed types based on the distribution of the support of the first itemset. The support of the itemset is the ratio of the number of transactions containing the itemset to the total number of transactions in the transaction dataset. The target frequent itemset determination module is configured to, for each watershed type, calculate the adaptive minimum support threshold corresponding to the second itemset based on the initial minimum support threshold and the length of the second itemset, and filter the target frequent itemsets corresponding to each watershed type based on the adaptive minimum support threshold, wherein the itemset length is the number of items contained in the itemset; The initial minimum confidence threshold determination module is configured to split the target frequent itemset into antecedent and consequent items for each watershed type, generate a first candidate association rule set, and determine the initial minimum confidence threshold based on the confidence distribution of the candidate association rules of a second preset length in the first candidate association rule set. The confidence is the ratio of the number of transactions that simultaneously contain antecedent and consequent items to the number of transactions that contain antecedent items. The adaptive minimum confidence threshold determination module is configured to calculate the adaptive minimum confidence threshold of the second candidate association rule set based on the initial minimum confidence threshold and the antecedent length for each watershed type and the different antecedent lengths. The flood disaster analysis module is configured to, for each watershed type, filter the second candidate association rules in the second candidate association rule set according to the adaptive minimum confidence threshold of each second candidate association rule set, to obtain the target association rule corresponding to each watershed type, so as to perform flood disaster risk analysis based on the target association rule.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1 to 7.

10. An electronic device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

Images