Meteorological early warning information pushing method and device based on user behavior data and flash message

Abstract

The invention provides a meteorological early warning information pushing method and device based on user behavior data and flash information, and relates to the field of big data analysis, and the method comprises the steps: determining meteorological disaster related information of a plurality of sub-regions included in a target region, and determining a meteorological early warning region of the target region in combination with the meteorological disaster data of the target region; determining a plurality of early warning user sides according to the meteorological early warning area of the target area; acquiring behavior data of a plurality of early warning user terminals, and dividing the plurality of early warning user terminals into a plurality of early warning user terminal groups; determining a target meteorological disaster early warning template according to the meteorological disaster data of the target area, and generating standard meteorological early warning information in combination with the meteorological disaster data of the target area; the meteorological early warning information corresponding to the early warning user terminal group is generated according to the behavior data of the early warning user terminals included in the early warning user terminal group and the standard meteorological early warning information, and a flash message is sent to the early warning user terminals included in the early warning user terminal group, so that the meteorological early warning information can be accurately and effectively pushed.

Description

[0001] Priority Statement

[0002] This application claims priority to Chinese application No. 202510390492.8, filed on March 31, 2025, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This invention relates to the field of big data analytics, and in particular to a method and device for pushing weather warning information based on user behavior data and flash messages. Background Technology

[0004] Meteorological disasters are frequent and account for more than 70% of natural disasters, often causing significant losses to economic and social development and people's lives and property. While humanity currently lacks the means to completely prevent or eliminate disasters, effective preventative measures can be taken to minimize the damage. Therefore, the timely delivery of meteorological disaster early warning information is crucial.

[0005] Currently, weather warnings are mainly issued via SMS. While this method is fast, it has the limitation that users need to actively click on the message to view the content. This means that in highly urgent situations, the information may not reach users in time, thus affecting the speed of emergency response.

[0006] Therefore, there is a need to provide methods and equipment for pushing meteorological warning information based on user behavior data and flash messages, so as to achieve accurate and effective delivery of meteorological warning information. Summary of the Invention

[0007] This invention provides a method for pushing meteorological warning information based on user behavior data and flash messages, comprising: determining meteorological disaster association information of multiple sub-regions including a target area; acquiring meteorological disaster data of the target area; determining a meteorological warning area of ​​the target area based on the meteorological disaster data of the target area and the meteorological disaster association information of the multiple sub-regions including the target area; determining multiple warning user terminals based on the meteorological warning area of ​​the target area; acquiring behavior data of the multiple warning user terminals and dividing the multiple warning user terminals into multiple warning user terminal groups; determining a target meteorological disaster warning template based on the meteorological disaster data of the target area; generating standard meteorological warning information based on the meteorological disaster data of the target area and the target meteorological disaster warning template; for each warning user terminal group, generating meteorological warning information corresponding to the warning user terminal group based on the behavior data of the warning user terminals included in the warning user terminal group and the standard meteorological warning information; and sending flash messages to the warning user terminals included in the warning user terminal group based on the meteorological warning information corresponding to the warning user terminal group.

[0008] Further, determining the meteorological disaster association information of multiple sub-regions included in the target area includes: obtaining historical meteorological disaster information of the target area; for each meteorological disaster, calculating the disaster co-occurrence coefficient of any two sub-regions of the target area based on the historical meteorological disaster information of the target area, and determining multiple first sub-region units corresponding to the meteorological disaster based on the disaster co-occurrence coefficient of any two sub-regions of the target area, each first sub-region unit including two sub-regions; obtaining historical behavior data of user terminals in the target area, wherein the historical behavior data includes the location of the user terminal at multiple historical time points obtained with user authorization; calculating the user movement association coefficient of any two sub-regions of the target area based on the historical behavior data of user terminals in the target area, and determining multiple second sub-region units based on the user movement association coefficient of any two sub-regions, each second sub-region unit including two sub-regions, wherein the meteorological disaster association information of the multiple sub-regions included in the target area includes at least multiple first sub-region units and multiple second sub-region units corresponding to each meteorological disaster.

[0009] Furthermore, based on meteorological disaster data of the target area and meteorological disaster association information of multiple sub-regions included in the target area, the meteorological warning area of ​​the target area is determined, including: for each sub-region unit, generating multiple training samples corresponding to the first sub-region unit based on historical meteorological disaster information of the target area, establishing an association prediction model corresponding to the first sub-region unit, and training the association prediction model corresponding to the first sub-region unit through the multiple training samples; determining the target meteorological disaster and key sub-regions based on the meteorological disaster data of the target area; and determining the meteorological warning area of ​​the target area based on the meteorological disaster, key sub-regions, meteorological disaster association information of multiple sub-regions included in the target area, and the association prediction model corresponding to each first sub-region unit.

[0010] Further, based on the meteorological disaster association information of the target meteorological disaster, key sub-regions, and multiple sub-regions included in the target area, as well as the association prediction model corresponding to each first sub-region unit, the meteorological warning area of ​​the target area is determined, including: determining the target first sub-region unit based on the multiple sub-region units and key sub-regions corresponding to the target meteorological disaster; predicting the disaster co-occurrence probability of the target first sub-region unit based on the historical meteorological disaster information of the target area using the association prediction model corresponding to the target first sub-region unit; determining the first supplementary sub-region based on the disaster co-occurrence probability of each target first sub-region unit; determining the second supplementary sub-region based on the multiple second sub-region units and key sub-regions; and determining the meteorological warning area of ​​the target area based on the key sub-region, the first supplementary sub-region, and the second supplementary sub-region.

[0011] Furthermore, based on the meteorological warning area of ​​the target area, multiple warning user terminals are identified, including: for each user terminal in the target area, based on the user terminal's behavioral data, calculating the first location association parameter between the user terminal and the meteorological warning area of ​​the target area, and determining whether the user terminal is a warning user terminal based on the first location association parameter between the user terminal and the meteorological warning area of ​​the target area.

[0012] Furthermore, behavioral data from multiple early warning user terminals are acquired, and these multiple early warning user terminals are divided into multiple early warning user terminal groups, including: for each early warning user terminal, determining the displacement activity coefficient of the early warning user terminal based on its behavioral data; using a K-Means clustering algorithm, dividing the multiple early warning user terminals into at least one early warning user terminal cluster based on the displacement activity coefficient of each early warning user terminal; for each early warning user terminal cluster, determining the second positional association parameter between the early warning user terminal and each sub-region included in the meteorological early warning area of ​​the target area based on the behavioral data of the early warning user terminals included in the cluster; using a K-Means clustering algorithm, dividing the multiple early warning user terminals included in the cluster into at least one early warning user terminal group based on the second positional association parameter between the early warning user terminals included in the cluster and each sub-region included in the meteorological early warning area of ​​the target area.

[0013] Further, based on the meteorological disaster data of the target area, a meteorological disaster early warning template is determined, including: acquiring multiple meteorological disaster early warning templates and meteorological disaster characteristics corresponding to each meteorological disaster early warning template, wherein the meteorological disaster characteristics include at least the meteorological disaster type and meteorological disaster level; acquiring current meteorological disaster characteristics based on the meteorological disaster data of the target area; and determining a target meteorological disaster early warning template from the multiple meteorological disaster early warning templates based on the current meteorological disaster characteristics and the meteorological disaster characteristics corresponding to each meteorological disaster early warning template.

[0014] Furthermore, based on the meteorological disaster data of the target area and the target meteorological disaster early warning template, standard meteorological early warning information is generated, including: obtaining key early warning information based on the meteorological disaster data of the target area, wherein the key early warning information includes at least the meteorological disaster type, meteorological disaster level, expected impact time, and key sub-areas; and generating standard meteorological early warning information based on the key early warning information and the target meteorological disaster early warning template.

[0015] Further, based on the behavioral data of the warning user terminals included in the warning user terminal group and the standard meteorological warning information, meteorological warning information corresponding to the warning user terminal group is generated, including: determining the meteorological warning style and response suggestions corresponding to the warning user terminal group based on the displacement activity coefficient of the warning user terminals included in the warning user terminal group; determining the warning priority of each sub-region of the meteorological warning area of ​​the target area of ​​the corresponding warning user terminal group based on the second position association parameter between the warning user terminals included in the warning user terminal group and each sub-region of the meteorological warning area of ​​the target area; and adjusting the standard meteorological warning information according to the meteorological warning style and response suggestions corresponding to the warning user terminal group and the warning priority of each sub-region of the meteorological warning area of ​​the target area of ​​the corresponding warning user terminal group through a style adjustment model, thereby generating the meteorological warning information corresponding to the warning user terminal group.

[0016] This invention provides a weather warning information push device based on user behavior data and flash SMS, applying the aforementioned weather warning information push method based on user behavior data and flash SMS, including: a region analysis module for determining meteorological disaster association information of multiple sub-regions included in the target region; a data acquisition module for acquiring meteorological disaster data of the target region; a region determination module for determining the weather warning area of ​​the target region based on the meteorological disaster data of the target region and the meteorological disaster association information of the multiple sub-regions included in the target region; and a user analysis module for determining multiple warning user terminals based on the meteorological warning area of ​​the target region; the user analysis module is further used to acquire the meteorological disaster data of the multiple warning user terminals. The system uses behavioral data to divide the multiple early warning user terminals into multiple early warning user terminal groups; a template determination module is used to determine a target meteorological disaster early warning template based on the meteorological disaster data of the target area; a meteorological early warning module is used to generate standard meteorological early warning information based on the meteorological disaster data of the target area and the target meteorological disaster early warning template; the meteorological early warning module is also used to generate meteorological early warning information corresponding to each early warning user terminal group based on the behavioral data of the early warning user terminals included in the early warning user terminal group and the standard meteorological early warning information, and to send flash messages to the early warning user terminals included in the early warning user terminal group based on the meteorological early warning information corresponding to the early warning user terminal group.

[0017] Compared with existing technologies, the weather warning information push method and device based on user behavior data and flash messages provided by this invention have at least the following beneficial effects:

[0018] 1. Flash notifications have a mandatory reminder function; users must manually click to confirm before closing, ensuring effective reception of warning information. Through precise target area division and meteorological disaster data correlation analysis, warning areas and user terminals can be more accurately determined, thereby reducing false alarms and missed alarms. Based on meteorological disaster data of the target area and meteorological disaster correlation information of sub-areas, specific meteorological warning areas are determined, achieving refined warnings. Based on user terminal behavior data, users are divided into different groups, and corresponding meteorological warning information is pushed to the groups, enhancing the personalization of warning information. Personalized warning information can better attract user attention and increase user awareness and importance of warning information. Timely and accurate meteorological warning information can help users prepare for disaster prevention and mitigation in advance, such as relocation and resettlement, and material reserves. Through effective warnings and response measures, casualties and property losses caused by meteorological disasters can be significantly reduced. Long-term and frequent meteorological warning information pushes can enhance public awareness of disaster prevention and mitigation and improve the overall disaster prevention capabilities of society.

[0019] 2. By calculating the co-occurrence coefficient of meteorological disasters between any two sub-regions of the target area, it is possible to identify which sub-regions have historically been prone to the same or similar meteorological disasters, thus providing an important basis for determining the warning area. The calculation of the user mobility correlation coefficient takes into account the movement patterns of users between different sub-regions, helping to more accurately locate user groups that may be affected by meteorological disasters. Using historical meteorological disaster information to generate training samples and establish a correlation prediction model, the co-occurrence probability of meteorological disasters between different sub-regions can be predicted, further improving the accuracy of the warning area. By comprehensively considering the target meteorological disaster, key sub-regions, meteorological disaster correlation information, and the results of the correlation prediction model, the warning area can be more finely divided, ensuring that warning information is only sent to users who are truly likely to be affected.

[0020] 3. Based on meteorological disaster data for the target area, the system automatically selects the most suitable template from multiple meteorological disaster warning templates. This ensures that the format and content of the warning information match the current meteorological disaster characteristics, improving the readability and usability of the information. Building upon the generated standard meteorological warning information, the system dynamically adjusts the warning information based on the behavioral data and location-related parameters of the warning user groups. This includes adjusting the warning style, response suggestions, and warning priorities for different sub-regions, making the warning information more aligned with user needs and actual circumstances. Attached Figure Description

[0021] This specification will be further described in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, the same numbers represent the same structures, wherein:

[0022] Figure 1 This is a schematic diagram of the meteorological warning area for the target area according to some embodiments of this specification;

[0023] Figure 2 This is a schematic diagram of the flash message corresponding to user terminal group A according to some embodiments of this specification;

[0024] Figure 3 This is a schematic diagram of the flash message corresponding to user terminal group B according to other embodiments of this specification;

[0025] Figure 4 This is a schematic diagram of a weather warning information push device based on user behavior data and flash messages, as shown in some embodiments of this specification. Detailed Implementation

[0026] To more clearly illustrate the technical solutions of the embodiments of this specification, the following briefly describes the drawings required for describing the embodiments. Obviously, the drawings described below are merely examples or embodiments of this specification. Those skilled in the art can apply this specification to other similar scenarios based on these drawings without inventive effort. Unless otherwise apparent from the context or otherwise noted, the same reference numerals in the figures represent the same structure or operation.

[0027] The method for pushing weather warning information based on user behavior data and flash messages may include the following process.

[0028] Step 110: Determine the meteorological disaster association information of multiple sub-regions included in the target area.

[0029] In some embodiments, step 110 specifically includes:

[0030] Obtain historical meteorological disaster information for the target area;

[0031] For each type of meteorological disaster, based on the historical meteorological disaster information of the target area, the disaster co-occurrence coefficient of any two sub-regions of the target area is calculated. Based on the disaster co-occurrence coefficient of any two sub-regions of the target area, multiple first sub-region units corresponding to the meteorological disaster are determined. Each first sub-region unit includes two sub-regions.

[0032] Obtain historical behavioral data of user terminals in the target area. The historical behavioral data includes the location of the user terminal at multiple historical time points obtained with the user's authorization. Ensure that the user's authorization has been obtained when obtaining the historical behavioral data of the user terminal, and comply with relevant privacy policies and regulations. User terminals whose numbers belong to the target area or have been located in the target area can be used as user terminals in the target area.

[0033] Based on the historical behavior data of users in the target area, calculate the user movement correlation coefficient between any two sub-regions of the target area. Based on the user movement correlation coefficient between any two sub-regions, determine multiple second sub-region units. Each second sub-region unit includes two sub-regions. The meteorological disaster correlation information of the multiple sub-regions included in the target area includes at least multiple first sub-region units and multiple second sub-region units corresponding to each type of meteorological disaster.

[0034] Specifically, historical meteorological disaster information for the target area can include relevant information on meteorological disasters that occurred in the target area over multiple historical time periods. For example, relevant information on meteorological disasters occurring over historical time periods may include:

[0035] I. Types of Meteorological Disasters

[0036] Drought: Records the time, scope, duration, and impact of drought disasters that have occurred in history. For example, a region experienced a severe drought in a certain year, leading to reduced crop yields and depletion of water resources.

[0037] Heavy Rain and Floods: Detailed records of the timing, rainfall, extent, and impact of heavy rain disasters. This includes information on river overflows, urban flooding, and flash floods caused by heavy rain.

[0038] Typhoon: Records the time, location, wind speed, path, and affected area of ​​a typhoon. Includes the disasters brought by the typhoon, such as strong winds, torrential rain, and storm surge.

[0039] Low-temperature freezing damage and snow disasters: Record the timing, scope, and severity of disasters such as low temperatures, frost, and snow. For example, a region suffered a severe snow disaster in a certain winter, resulting in traffic disruptions and crop damage.

[0040] Wind and hail disasters: Records the time, extent, wind force level, and hail size of disasters such as strong winds and hail. Includes the impact of wind and hail disasters on crops, buildings, and personnel safety.

[0041] II. Impact of the Disaster

[0042] Agricultural losses: Assessing the impact of meteorological disasters on crop yield, quality, and market supply. This includes affected area, degree of yield reduction, and loss of crop value.

[0043] Economic losses: Statistics include direct and indirect economic losses caused by meteorological disasters. Direct economic losses include crop losses, building damage, and infrastructure damage; indirect economic losses include losses due to production stoppages and business interruptions, and transportation disruptions.

[0044] Casualties: Record the casualties caused by meteorological disasters, including the number of deaths, injuries, and missing persons.

[0045] III. Disaster Response and Rescue

[0046] Early warning and forecast: Record disaster early warning and forecast information issued by meteorological departments, as well as the effectiveness and accuracy of early warning equipment.

[0047] Emergency Response: Describes the emergency response measures taken by the government and relevant departments after a disaster occurs, including rescue operations, resource allocation, and personnel evacuation.

[0048] Post-disaster recovery: Documenting the progress and achievements of post-disaster recovery efforts, including rebuilding homes, repairing infrastructure, and restoring agricultural production.

[0049] The target area can be divided into multiple sub-regions according to administrative management needs. For example, a city can be divided into multiple districts, and a county into multiple towns.

[0050] For each sub-region of the target area, the status of the sub-region in multiple historical time periods (e.g., one week, one month) can be coded based on the historical meteorological disaster information of the target area. Taking drought as an example, the drought status code of the sub-region in a historical time period can be calculated based on whether drought occurred in the sub-region in a certain historical time period, the severity of the drought, and the duration of the drought. For example, the coding method for whether drought occurred is: 0: no drought occurred, 1: drought occurred (regardless of severity or duration). The coding method for drought severity is: 1: mild drought (e.g., soil moisture is below the threshold but crops are not significantly affected); 2: moderate drought (e.g., crop yield reduction of 10%-30%); 3: severe drought (e.g., crop failure, water depletion). The coding method for drought duration is direct counting: the duration is recorded in "days" or "time periods". If sub-region A experiences drought in historical time period 2, and the severity is moderate with a duration of 15 days, then the drought status code of sub-region A in historical time period 2 is 1215. If sub-region A did not experience drought in historical time period 3, then the drought status code for sub-region A in historical time period 3 is 0. The drought co-occurrence coefficients for the two sub-regions can be calculated using correlation coefficient formulas (e.g., Spearman rank correlation coefficient) based on their drought status codes across multiple historical time periods. Two sub-regions whose corresponding meteorological disaster co-occurrence coefficients are greater than a disaster co-occurrence coefficient threshold (e.g., 0.6) can be considered as a first sub-region unit.

[0051] For any two sub-regions, the number of user terminals that moved between the two sub-regions during multiple historical time periods can be determined based on the historical behavior data of user terminals in the target region. The user movement correlation coefficient between the two sub-regions is then calculated based on the number of user terminals that moved between the two sub-regions during each historical time period.

[0052] Multiple historical time points can be divided into multiple historical time periods. For any two sub-regions, the number of bidirectional users flowing between the two sub-regions in each historical time period (e.g., one day, one week) can be determined based on the historical behavior data of users in the target region. The number of bidirectional users can include the number of users flowing from one sub-region to another and the number of users flowing from one sub-region to another. For example, for sub-regions A and B, the number of bidirectional users flowing between sub-regions A and B in a certain historical time period can include the number of users flowing from sub-region A to sub-region B and the number of users flowing from sub-region B to sub-region A. It is understandable that if a user moves from sub-region A to sub-region B and then from sub-region B to sub-region A within a historical time period, the number of users flowing from sub-region A to sub-region B increases by 1, and the number of users flowing from sub-region B to sub-region A also increases by 1.

[0053] The user mobility correlation coefficient between two sub-regions can be calculated using correlation coefficient formulas (e.g., Pearson correlation coefficient, Spearman rank correlation coefficient, etc.) based on the number of bidirectional users in each historical time period. Two sub-regions whose user mobility correlation coefficients are greater than a threshold (e.g., 0.6) can be considered as a second sub-region unit.

[0054] Understandably, by encoding the historical disaster status of sub-regions (such as the severity and duration of drought) and calculating disaster co-occurrence coefficients (such as the Spearman rank correlation coefficient), the synchronicity of disaster occurrence in different sub-regions can be quantified. If the drought co-occurrence coefficient between sub-regions A and B is 0.8 (> the threshold of 0.6), it indicates that drought often occurs simultaneously or successively in both regions. In early warning, A and B should be treated as a single unit (the first sub-region unit) to avoid information fragmentation due to independent warnings. This reduces early warning blind spots and is particularly suitable for the early identification of disaster chains (such as heavy rain causing urban flooding, which then spreads to downstream areas).

[0055] Traditional early warning systems rely on geographical proximity or administrative boundaries to delineate areas, easily overlooking functional connections (such as between commercial and residential areas). By calculating user mobility correlation coefficients, data can replace empirical judgments, reducing human bias. Combining disaster co-occurrence coefficients and user mobility correlation coefficients, a dual correlation network between disaster propagation and user movement can be constructed, providing multi-dimensional evidence for meteorological early warning information delivery.

[0056] Step 120: Obtain meteorological disaster data for the target area.

[0057] Specifically, meteorological disaster data for the target area can be information related to predicted meteorological disasters that will occur in the target area over a future period. This includes, for example, the type of meteorological disaster, its duration, severity, and location.

[0058] Step 130: Determine the meteorological warning area of ​​the target area based on the meteorological disaster data of the target area and the meteorological disaster association information of the multiple sub-regions included in the target area.

[0059] In some embodiments, step 130 specifically includes: for each sub-regional unit, generating multiple training samples corresponding to the first sub-regional unit based on historical meteorological disaster information of the target area, establishing an association prediction model corresponding to the first sub-regional unit, and training the association prediction model corresponding to the first sub-regional unit through the multiple training samples corresponding to the first sub-regional unit. The association prediction model can be a long short-term memory network model, the input of the association prediction model includes meteorological data of the target area, and the output of the association prediction model includes the disaster co-occurrence probability of the meteorological disaster occurring in the first sub-regional unit.

[0060] Based on meteorological disaster data of the target area, target meteorological disasters and key sub-regions are identified. Target meteorological disasters are meteorological disasters that will occur in the target area as indicated by the meteorological disaster data, and key sub-regions are sub-regions in the target area indicated by the meteorological disaster data that have been identified as having meteorological disasters. Meteorological disaster data of the target area can be obtained from external data sources.

[0061] Based on the meteorological disaster information of the target meteorological disaster, key sub-regions, and multiple sub-regions included in the target region, as well as the correlation prediction model corresponding to each first sub-region unit, the meteorological warning area of ​​the target region is determined.

[0062] Specifically, association prediction models can include:

[0063] 1. Input layer

[0064] Meteorological data: precipitation, temperature, wind speed, humidity, air pressure, etc. (select key elements according to the type of disaster).

[0065] Input dimensions: (time step, number of features), for example, daily precipitation, temperature, and disaster labels (number of features = 3) over the past 30 days (time step = 30).

[0066] 2. LSTM layer

[0067] Single-layer or double-layer LSTM, each layer contains 64-256 hidden units.

[0068] Activation function: tanh or ReLU.

[0069] Add Dropout (e.g., 0.2-0.5) after the LSTM layer to prevent overfitting.

[0070] 3. Fully Connected Layer

[0071] After flattening the output of the LSTM, it is mapped to the final output through 1-2 fully connected layers (such as 128 neurons).

[0072] Activation function: sigmoid.

[0073] 4. Output layer

[0074] Output the co-occurrence probability of this meteorological disaster occurring in the first sub-region unit.

[0075] The loss function for training the association prediction model can be the cross-entropy loss function. The learning rate of the Adam optimizer is set to 0.001. Evaluation metrics can include F1 score, confusion matrix, etc. The training set (70%), validation set (15%), and test set (15%) are divided chronologically. Mini-batch gradient descent is performed using a batch size (e.g., 32-128). Training stops if the validation loss does not decrease for five consecutive rounds. The model with the best performance on the validation set is saved.

[0076] In some embodiments, the meteorological warning area of ​​the target area is determined based on the meteorological disaster association information of the target meteorological disaster, key sub-regions, multiple sub-regions included in the target area, and the association prediction model corresponding to each first sub-region unit, including:

[0077] Based on the multiple sub-regional units and key sub-regions corresponding to the target meteorological disaster, the first sub-regional unit of the target is determined. For example, the first sub-regional unit in which one of the two included sub-regions is a key sub-region and the other is not a key sub-region is taken as the first sub-regional unit of the target.

[0078] Based on the historical meteorological disaster information of the target area, the probability of disaster co-occurrence in the first sub-region of the target area is predicted by the correlation prediction model corresponding to the first sub-region of the target area.

[0079] Based on the disaster co-occurrence probability of each target first sub-region unit, a first supplementary sub-region is determined. For example, the target first sub-region unit with a disaster co-occurrence probability greater than the disaster co-occurrence probability threshold (e.g., 60%) is taken as a supplementary first sub-region unit, and the sub-regions that are not critical sub-regions in the supplementary first sub-region units are taken as the first supplementary sub-regions.

[0080] Based on multiple second sub-region units and key sub-regions, a second supplementary sub-region is determined. For example, a second sub-region unit in which one of the two included sub-regions is a key sub-region and the other is not a key sub-region is taken as a target second sub-region unit, and the sub-region in the target second sub-region unit that is not a key sub-region is taken as a second supplementary sub-region.

[0081] Based on the key sub-region, the first supplementary sub-region, and the second supplementary sub-region, the meteorological warning area of ​​the target region is determined. For example, the union of the first and second supplementary sub-regions is taken, and the union of the first and second supplementary sub-regions with the key sub-region is taken to determine the meteorological warning area of ​​the target region. Figure 1 As shown, the area within the red dashed box is the weather warning area for the target region in the example.

[0082] Understandably, traditional methods rely on geographical proximity or administrative boundaries, easily overlooking user mobility risks (such as commuting routes crossing disaster areas). This method quantifies the intensity of disaster associations between sub-regions through disaster co-occurrence coefficients and captures the coupling relationship between population flow and disasters by combining user mobility association coefficients, accurately identifying potential risk networks across functional zones (such as residential areas-commercial areas-transportation hubs). It integrates disaster propagation patterns with user behavior patterns to form a dual "disaster-user" association network, identifying risk areas, replacing traditional experience-based judgments, and reducing subjective bias. Compared to traditional "full coverage" early warning, this method narrows the warning scope to high-risk associated areas through dynamic boundary delineation. Experimental data shows that, while maintaining the same recall rate, it can reduce redundant early warning transmissions by more than 30%, reducing system load and user information overload.

[0083] Step 140: Determine multiple early warning user terminals based on the meteorological early warning area of ​​the target region.

[0084] In some embodiments, step 140 specifically includes:

[0085] For each user terminal in the target area, the first location association parameter between the user terminal and the meteorological warning area of ​​the target area is calculated based on the user terminal's behavior data. Based on the first location association parameter between the user terminal and the meteorological warning area of ​​the target area, it is determined whether the user terminal is a warning user terminal.

[0086] Specifically, based on user behavior data, the location of the user at multiple points in time within the current period (e.g., the last 3 months) can be determined.

[0087] Based on the location of the user terminal at multiple time points in the current period (e.g., the last 3 months), a location sequence of the user terminal is generated. Each element of the location sequence represents the location status of the user terminal at a certain time point. If the user terminal is located within the weather warning area of ​​the target region at a certain time point, the value of the element is 2; otherwise, the value of the element is 1.

[0088] The first location correlation parameter between the user terminal and the meteorological warning area of ​​the target region is calculated based on the user terminal's location sequence using the following formula:

[0089]

[0090] Where, γ (i,1) V is the first location association parameter between the i-th user terminal and the meteorological warning area of ​​the target area. (i,t) Let t be the value of the t-th element in the position sequence, where T is the total number of sampling points in the current period.

[0091] If the first location association parameter of the user terminal and the meteorological warning area of ​​the target area is greater than the first location association parameter threshold, the user terminal is determined to be a warning user terminal.

[0092] Step 150: Obtain behavioral data from multiple warning user terminals and divide the multiple warning user terminals into multiple warning user terminal groups.

[0093] In some embodiments, step 150 specifically includes:

[0094] For each early warning user terminal, the displacement activity coefficient of the early warning user terminal is determined based on the behavioral data of the early warning user terminal;

[0095] Using the K-Means clustering algorithm, multiple early warning user terminals are divided into at least one early warning user terminal cluster based on the displacement activity coefficient of each early warning user terminal;

[0096] For each early warning user terminal cluster, based on the behavioral data of the early warning user terminals included in the cluster, the second location association parameter between the early warning user terminal and each sub-region included in the meteorological early warning area of ​​the target region is determined. Using the K-Means clustering algorithm, based on the second location association parameter between the early warning user terminals included in the cluster and each sub-region included in the meteorological early warning area of ​​the target region, the multiple early warning user terminals included in the cluster are divided into at least one early warning user terminal group.

[0097] Specifically, for each warning user terminal, the location of the warning user terminal at multiple time points in the current period (e.g., the last 3 months) can be encoded. For example, if the warning user terminal is located in sub-region A of the target area at the first time point, its location is encoded as 1; if it is located in sub-region B of the target area at the second time point, its location is encoded as 2; and if it is located outside the target area at the nth time point, its location is encoded as K+1, where K is the total number of sub-regions included in the target area.

[0098] The displacement activity coefficient of the early warning user terminal can be calculated using the following formula:

[0099]

[0100] Where, σ i Let E be the displacement activity coefficient of the i-th early warning user terminal. (i,t)The location code for the i-th warning user terminal at the t-th time point in the current period is given, where T is the total number of time points sampled in the current period.

[0101] The method for determining the second location association parameter of each sub-region included in the meteorological warning area between the user terminal and the target area is similar to the method for determining the first location association parameter of the meteorological warning area between the user terminal and the target area, and will not be repeated here.

[0102] For each warning user terminal, a parameter vector corresponding to the warning user terminal can be generated based on the second location association parameters between the warning user terminal and each sub-region included in the meteorological warning area of ​​the target area. For example, the parameter vector corresponding to warning user terminal A is (γ (i,1,2) γ (i,2,2) …γ (i,f,2) ), where γ (i,1,2) γ is the second location association parameter between the i-th early warning user terminal and the first sub-region included in the meteorological early warning area of ​​the target area. (i,2,2) γ is the second location association parameter between the i-th warning user terminal and the second sub-region included in the meteorological warning area of ​​the target area. (i,f,2) The second location association parameter is the relationship between the i-th warning user terminal and the f-th sub-region included in the meteorological warning area of ​​the target area, where f is the total number of sub-regions included in the meteorological warning area of ​​the target area.

[0103] For any two early warning user terminals included in an early warning user terminal cluster, the cosine distance between the parameter vectors corresponding to the two early warning user terminals can be calculated as the clustering distance between the two early warning user terminals. Using the K-Means clustering algorithm, based on the clustering distance between any two early warning user terminals included in the early warning user terminal cluster, the multiple early warning user terminals included in the early warning user terminal cluster are divided into at least one early warning user terminal group.

[0104] Understandably, quantifying differences in user movement patterns using displacement activity coefficients and refining regional correlation strength using second location correlation parameters provides multidimensional and objective evidence for grouping. Employing two-stage K-Means clustering significantly reduces the complexity of high-dimensional data clustering and improves processing speed for large-scale user terminals. The hierarchical framework ensures that users within the same group share similar movement activity and regional correlation patterns, avoiding biases caused by grouping based on a single indicator and improving the matching accuracy of early warning information.

[0105] Step 160: Determine the target meteorological disaster early warning template based on the meteorological disaster data of the target area.

[0106] In some embodiments, step 160 specifically includes:

[0107] Obtain multiple meteorological disaster warning templates and the meteorological disaster characteristics corresponding to each meteorological disaster warning template. The meteorological disaster characteristics include at least the meteorological disaster type and the meteorological disaster level.

[0108] Based on meteorological disaster data for the target area, obtain the current characteristics of meteorological disasters;

[0109] Based on the current meteorological disaster characteristics and the meteorological disaster characteristics corresponding to each meteorological disaster warning template, the target meteorological disaster warning template is determined from multiple meteorological disaster warning templates.

[0110] Specifically, meteorological disaster warning templates that match the type and level of meteorological disasters can be used as target meteorological disaster warning templates.

[0111] Step 170: Generate standard meteorological warning information based on meteorological disaster data of the target area and the target meteorological disaster warning template.

[0112] In some embodiments, step 170 specifically includes:

[0113] Based on meteorological disaster data of the target area, obtain key early warning information, which includes at least the type of meteorological disaster, the level of meteorological disaster, the expected impact time, and key sub-regions. For example, meteorological disaster type (such as rainstorm, typhoon, drought, etc.), meteorological disaster level (such as level 1, level 2, level 3, etc.), expected impact time (such as from XX month XX day XX hour to XX month XX day XX hour), and key sub-regions (such as XX city XX district, XX county, etc.).

[0114] Standard meteorological warning information is generated based on key early warning information and target meteorological disaster early warning templates.

[0115] Specifically, the standard format for early warning information includes a title, body, issuing unit, and release date. The target meteorological disaster early warning template contains placeholders for inserting key early warning information later. Fill the placeholders in the selected template with the key early warning information. For example, the title is generated based on the type and level of the meteorological disaster, such as "Blue Alert for Heavy Rain." The body describes in detail the type, level, expected impact time, and key sub-regions of the meteorological disaster, and may include disaster prevention and mitigation recommendations. The issuing unit is the name of the unit issuing the early warning information. The release date is the date the early warning information was released.

[0116] Step 180: For each warning user terminal group, generate meteorological warning information corresponding to the warning user terminal group based on the behavioral data of the warning user terminals included in the warning user terminal group and the standard meteorological warning information, and send flash messages to the warning user terminals included in the warning user terminal group based on the meteorological warning information corresponding to the warning user terminal group.

[0117] In some embodiments, step 180 specifically includes:

[0118] Based on the displacement activity coefficient of the warning user terminals included in the warning user terminal group, determine the meteorological warning style and response suggestions corresponding to the warning user terminal group;

[0119] Based on the second location association parameters of the warning user terminals included in the warning user terminal group and each sub-region included in the meteorological warning area of ​​the target area, the warning priority of each sub-region included in the meteorological warning area of ​​the target area of ​​the corresponding warning user terminal group is determined. For example, the average value of the second location association parameters of the warning user terminals included in the warning user terminal group and each sub-region included in the meteorological warning area of ​​the target area is calculated to obtain the average value of the second location association parameters of the warning user terminal group and the sub-region. The sub-region with a larger average value of the second location association parameters has a higher warning priority.

[0120] Using a style adjustment model, standard meteorological warning information is adjusted based on the meteorological warning style and response suggestions corresponding to the warning user group, and the warning priority of each sub-region included in the meteorological warning area of ​​the target area of ​​the corresponding warning user group, to generate meteorological warning information corresponding to the warning user group. The style adjustment model can be a Large Language Model (LLM). In the meteorological warning information corresponding to the warning user group, the warning prompts for sub-regions within the meteorological warning area of ​​the target area with higher warning priority are displayed earlier, for example, Figure 2 In the meteorological warning information corresponding to user group A shown in the warning diagram, the warning priority of sub-region A1 is higher than that of sub-region A2, and the warning priority of sub-region A2 is higher than that of sub-region A3. Figure 3 In the warning user terminal group B shown, the warning priority of sub-region A2 is higher than that of sub-region A1, and the warning priority of sub-region A1 is higher than that of sub-region A3.

[0121] Specifically, multiple thresholds can be set to divide the early warning user terminal groups into high displacement activity groups, medium displacement activity groups, and low displacement activity groups. For example, a first threshold (e.g., 0.2) and a second threshold (e.g., 0.4) can be set, where the first threshold is less than the second threshold. The average displacement activity coefficient of the early warning user terminals included in the early warning user terminal group is calculated to obtain the average displacement activity coefficient. The early warning user terminal group with an average displacement activity coefficient less than the first threshold is designated as the low displacement activity group, the early warning user terminal group with an average displacement activity coefficient greater than or equal to the first threshold and less than the second threshold is designated as the medium displacement activity group, and the early warning user terminal group with an average displacement activity coefficient greater than or equal to the second threshold is designated as the high displacement activity group.

[0122] Different weather warning styles and response recommendations are set for the high-displacement active group, the medium-displacement active group, and the low-displacement active group.

[0123] As examples only, the meteorological warning styles and response suggestions corresponding to the high-displacement-activity group, the medium-displacement-activity group, and the low-displacement-activity group can be as follows:

[0124] High displacement active group:

[0125] Warning style: Since users move frequently and may be in different geographical locations often, warning information should be concise and clear, highlighting key information, such as "Urgent! Heavy rain is about to occur in XX area. Please find a safe place immediately."

[0126] Recommendations: Users are advised to install a weather warning app and enable real-time push notifications to ensure they receive timely warnings regardless of their location. Users are also reminded to carry necessary emergency items such as raincoats and flashlights.

[0127] Medium-displacement active group:

[0128] Warning style: For this type of user, the warning information can be more detailed, including the type, level, expected time and range of the meteorological disaster, such as "Typhoon XX is expected to make landfall in XX city tonight, with wind force reaching level 12 or above. Please take precautions against the wind and reinforce your defenses."

[0129] Recommendations: Users are advised to pay attention to weather warnings issued by local meteorological departments and take preventative measures in advance, such as reinforcing doors and windows and clearing clutter from balconies. Users are also reminded to avoid dangerous areas to ensure personal safety.

[0130] Low displacement active group:

[0131] Warning style: Since users move less, they may pay more attention to long-term or regional weather changes. Warning information can be more in-depth, including the background of meteorological disasters, historical data, and possible impacts, such as "Recently, there has been frequent rainfall in XX area, and the soil moisture content is close to saturation, which may easily lead to secondary disasters such as landslides. Residents are advised to take precautions."

[0132] Understandably, the system dynamically matches early warning expressions and provides targeted response suggestions based on user group displacement activity coefficients to improve the acceptability of meteorological early warning information; it objectively determines the risk ranking of sub-regions by using the mean of the second location association parameters to ensure that early warning resources focus on high-exposure risk areas; it utilizes a large language model to integrate user behavior and regional risks to automatically generate structured and hierarchical early warning content, optimizing information readability and usability; and it displays sub-region prompts according to early warning priority, enabling users to quickly obtain relevant core risks and enhancing the timeliness of early warnings.

[0133] Figure 4 This is a schematic diagram of a weather warning information push device based on user behavior data and flash messages, as shown in some embodiments of this specification. Figure 4As shown, a weather warning information push device based on user behavior data and flash messages may include a regional analysis module, a data acquisition module, a regional determination module, a user analysis module, a template determination module, and a weather warning module.

[0134] The regional analysis module is used to determine the meteorological disaster association information of multiple sub-regions included in the target area;

[0135] The data acquisition module is used to acquire meteorological disaster data for the target area;

[0136] The region determination module is used to determine the meteorological warning area of ​​the target region based on the meteorological disaster data of the target region and the meteorological disaster association information of multiple sub-regions included in the target region;

[0137] The user analysis module is used to determine multiple early warning user terminals based on the meteorological early warning area of ​​the target region;

[0138] The user analysis module is also used to acquire behavioral data from multiple warning user terminals and divide these multiple warning user terminals into multiple warning user terminal groups.

[0139] The template determination module is used to determine the target meteorological disaster early warning template based on the meteorological disaster data of the target area;

[0140] The meteorological early warning module is used to generate standard meteorological early warning information based on meteorological disaster data of the target area and the target meteorological disaster early warning template;

[0141] The weather warning module is also used to generate weather warning information corresponding to each warning user terminal group based on the behavioral data of the warning user terminals included in the warning user terminal group and standard weather warning information, and to send flash messages to the warning user terminals included in the warning user terminal group based on the weather warning information corresponding to the warning user terminal group.

[0142] The weather warning information push device based on user behavior data and flash message can be used to execute the weather warning information push method based on user behavior data and flash message, which will not be elaborated here.

[0143] Finally, it should be understood that the embodiments described in this specification are intended only to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, by way of example and not limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly described and illustrated in this specification.

Claims

1. A method for pushing weather warning information based on user behavior data and flash messages, characterized in that, include: Determine the meteorological disaster association information of multiple sub-regions included in the target area; Obtain meteorological disaster data for the target area; Based on meteorological disaster data of the target area and meteorological disaster correlation information of multiple sub-regions included in the target area, the meteorological warning area of ​​the target area is determined; Based on the meteorological warning areas of the target region, multiple warning user terminals are determined; Acquire behavioral data from multiple warning user terminals and divide the multiple warning user terminals into multiple warning user terminal groups; Based on the meteorological disaster data of the target area, determine the target meteorological disaster early warning template; Based on the meteorological disaster data of the target area and the target meteorological disaster early warning template, standard meteorological early warning information is generated; For each warning user terminal group, based on the behavioral data of the warning user terminals included in the warning user terminal group and the standard meteorological warning information, a meteorological warning information corresponding to the warning user terminal group is generated, and a flash message is sent to the warning user terminals included in the warning user terminal group based on the meteorological warning information corresponding to the warning user terminal group.

2. The method for pushing meteorological early warning information based on user behavior data and flash messages according to claim 1, characterized in that, Determine meteorological disaster association information for multiple sub-regions included in the target area, including: Obtain historical meteorological disaster information for the target area; For each type of meteorological disaster, based on the historical meteorological disaster information of the target area, the disaster co-occurrence coefficient of any two sub-regions of the target area is calculated. Based on the disaster co-occurrence coefficient of any two sub-regions of the target area, multiple first sub-region units corresponding to the meteorological disaster are determined. Each first sub-region unit includes two sub-regions. Obtain historical behavior data of user terminals in the target area, wherein the historical behavior data includes the location of the user terminal at multiple historical time points obtained with user authorization; Based on the historical behavior data of users in the target area, the user movement correlation coefficient between any two sub-regions of the target area is calculated. Based on the user movement correlation coefficient between any two sub-regions, multiple second sub-region units are determined. Each second sub-region unit includes two sub-regions. The meteorological disaster correlation information of the multiple sub-regions included in the target area includes at least multiple first sub-region units and multiple second sub-region units corresponding to each meteorological disaster.

3. The method for pushing meteorological early warning information based on user behavior data and flash messages according to claim 2, characterized in that, Based on meteorological disaster data of the target area and meteorological disaster correlation information of multiple sub-regions included in the target area, the meteorological warning area of ​​the target area is determined, including: For each sub-regional unit, based on the historical meteorological disaster information of the target area, multiple training samples are generated corresponding to the first sub-regional unit, and an association prediction model corresponding to the first sub-regional unit is established. The association prediction model corresponding to the first sub-regional unit is trained using the multiple training samples corresponding to the first sub-regional unit. Based on meteorological disaster data of the target area, identify the target meteorological disasters and key sub-regions; Based on the meteorological disaster information of the target meteorological disaster, key sub-regions, multiple sub-regions included in the target region, and the correlation prediction model corresponding to each first sub-region unit, the meteorological warning area of ​​the target region is determined.

4. The method for pushing meteorological early warning information based on user behavior data and flash messages according to claim 3, characterized in that, Based on the meteorological disaster association information of the target meteorological disaster, key sub-regions, and multiple sub-regions included in the target region, and the association prediction model corresponding to each first sub-region unit, the meteorological warning area of ​​the target region is determined, including: Based on the multiple sub-regional units and key sub-regions corresponding to the target meteorological disaster, the first sub-regional unit of the target is determined; Based on the historical meteorological disaster information of the target area, the probability of disaster co-occurrence in the first sub-region of the target area is predicted by the correlation prediction model corresponding to the first sub-region of the target area. The first supplementary sub-region is determined based on the disaster co-occurrence probability of the first sub-region unit of each target; The second supplementary sub-region is determined based on multiple second sub-region units and key sub-regions; The meteorological warning area of ​​the target area is determined based on the key sub-region, the first supplementary sub-region, and the second supplementary sub-region.

5. The method for pushing meteorological early warning information based on user behavior data and flash messages according to claim 3, characterized in that, Based on the meteorological warning area of ​​the target region, multiple warning user terminals are identified, including: For each user terminal in the target area, the first location association parameter between the user terminal and the meteorological warning area of ​​the target area is calculated based on the user terminal's behavior data. Based on the first location association parameter between the user terminal and the meteorological warning area of ​​the target area, it is determined whether the user terminal is a warning user terminal.

6. The method for pushing meteorological early warning information based on user behavior data and flash messages according to claim 1, characterized in that, Acquire behavioral data from multiple early warning user terminals, and divide the multiple early warning user terminals into multiple early warning user terminal groups, including: For each early warning user terminal, the displacement activity coefficient of the early warning user terminal is determined based on the behavioral data of the early warning user terminal; Using the K-Means clustering algorithm, the multiple early warning user terminals are divided into at least one early warning user terminal cluster based on the displacement activity coefficient of each early warning user terminal; For each early warning user terminal cluster, based on the behavioral data of the early warning user terminals included in the cluster, the second location association parameter between the early warning user terminal and each sub-region included in the meteorological early warning area of ​​the target region is determined. Using the K-Means clustering algorithm, based on the second location association parameter between the early warning user terminals included in the cluster and each sub-region included in the meteorological early warning area of ​​the target region, the multiple early warning user terminals included in the cluster are divided into at least one early warning user terminal group.

7. The method for pushing meteorological early warning information based on user behavior data and flash messages according to any one of claims 1-6, characterized in that, Based on meteorological disaster data for the target area, a meteorological disaster early warning template is determined, including: Multiple meteorological disaster early warning templates and meteorological disaster characteristics corresponding to each meteorological disaster early warning template are obtained, wherein the meteorological disaster characteristics include at least the meteorological disaster type and the meteorological disaster level; Based on the meteorological disaster data of the target area, obtain the current meteorological disaster characteristics; Based on the current meteorological disaster characteristics and the meteorological disaster characteristics corresponding to each meteorological disaster warning template, the target meteorological disaster warning template is determined from multiple meteorological disaster warning templates.

8. The method for pushing meteorological early warning information based on user behavior data and flash messages according to any one of claims 1-6, characterized in that, Based on the meteorological disaster data of the target area and the target meteorological disaster early warning template, standard meteorological early warning information is generated, including: Based on meteorological disaster data of the target area, key early warning information is obtained, wherein the key early warning information includes at least the type of meteorological disaster, the level of meteorological disaster, the expected impact time, and key sub-regions; Based on the key early warning information and the target meteorological disaster early warning template, standard meteorological early warning information is generated.

9. The method for pushing meteorological early warning information based on user behavior data and flash messages according to claim 6, characterized in that, Based on the behavioral data of the warning user terminals included in the warning user terminal group and the standard meteorological warning information, meteorological warning information corresponding to the warning user terminal group is generated, including: Based on the displacement activity coefficient of the warning user terminals included in the warning user terminal group, determine the meteorological warning style and response suggestions corresponding to the warning user terminal group; Based on the second location association parameters between the early warning user terminals included in the early warning user terminal group and each sub-region included in the meteorological early warning area of ​​the target area, the early warning priority of each sub-region included in the meteorological early warning area of ​​the target area of ​​the corresponding early warning user terminal group is determined. The standard meteorological warning information is adjusted based on the meteorological warning style and response suggestions corresponding to the warning user group and the warning priority of each sub-region included in the meteorological warning area of ​​the target area of ​​the warning user group, using a style adjustment model to generate the meteorological warning information corresponding to the warning user group.

10. A weather warning information push device based on user behavior data and flash messages, characterized in that, The weather warning information push method based on user behavior data and flash message as described in any one of claims 1-9 includes: The regional analysis module is used to determine the meteorological disaster association information of multiple sub-regions included in the target area; The data acquisition module is used to acquire meteorological disaster data for the target area; The region determination module is used to determine the meteorological warning area of ​​the target region based on the meteorological disaster data of the target region and the meteorological disaster association information of multiple sub-regions included in the target region; The user analysis module is used to determine multiple early warning user terminals based on the meteorological early warning area of ​​the target region; The user analysis module is also used to acquire behavioral data from multiple warning user terminals and divide the multiple warning user terminals into multiple warning user terminal groups. The template determination module is used to determine the target meteorological disaster early warning template based on the meteorological disaster data of the target area; The meteorological early warning module is used to generate standard meteorological early warning information based on the meteorological disaster data of the target area and the target meteorological disaster early warning template; The weather warning module is also used to generate weather warning information corresponding to each warning user terminal group based on the behavioral data of the warning user terminals included in the warning user terminal group and standard weather warning information, and to send flash messages to the warning user terminals included in the warning user terminal group based on the weather warning information corresponding to the warning user terminal group.

Images