Method and system for property insurance claim risk management based on multi-source spatio-temporal precipitation data

Abstract

The present application relates to the technical field of insurance claim settlement, and discloses a property insurance claim settlement risk management method and system based on multi-source spatio-temporal precipitation data, which comprises the following steps: obtaining a precipitation data set of a target region in a target period; determining a precipitation time series stability index based on the continuity change characteristics of the precipitation data set in the time series dimension, and constructing a regional difference dispersion degree based on the distribution discrete characteristics of the precipitation data set in the spatial dimension; obtaining a risk spatial correlation degree according to the historical claim case region and the current precipitation grid, in combination with the intensity evolution mode of the precipitation data set; generating a claim settlement threshold dynamic correction coefficient by fusing the precipitation time series stability index, the regional difference dispersion degree and the risk spatial correlation degree; and dynamically adjusting a preset claim settlement trigger threshold based on the claim settlement threshold dynamic correction coefficient to form a claim settlement risk control management strategy, which significantly optimizes the survey resource allocation efficiency and the disaster prevention and loss reduction action efficiency of the insurance company under the premise of improving the claim settlement risk identification accuracy.

Description

Technical Field

[0001] This invention relates to the field of insurance claims technology, and more specifically, to a method and system for property insurance claims risk management based on multi-source spatiotemporal precipitation data. Background Technology

[0002] Property insurance claims risk management refers to the technical process of assessing property insurance payout risks and guiding the pre-deployment of claims resources by analyzing the correlation between meteorological disaster-causing factors and loss data of insured objects.

[0003] Current technologies primarily rely on single observation data from ground-based meteorological stations, combined with fixed claim trigger thresholds set by human experience, for risk assessment. Some solutions incorporate numerical weather prediction model outputs as supplementary references. However, existing technologies often face three major shortcomings: First, a single data source cannot capture the fine spatiotemporal evolution characteristics of precipitation, making it difficult to capture the complex nonlinear relationship between precipitation indicators and insurance claims. Second, static thresholds lack dynamic response capabilities to the stability and spatial variability of precipitation processes, easily leading to threshold failure during persistent heavy precipitation events. This results in subsequent risk level classifications relying heavily on statistical quantiles or subjective experience, lacking objective and quantitative risk identification based on the data itself. Third, a quantitative assessment mechanism that establishes a spatial correlation between current precipitation patterns and historical claim cases has not been established, making it difficult to predict the transmission path and impact range of disaster risks. These problems cause a disconnect between claims decisions and actual risk distribution, inefficient allocation of investigation resources, delayed early warning signals, and inefficient disaster prevention and mitigation actions by insurance companies. Summary of the Invention

[0004] This invention provides a property insurance claims risk management method and system based on multi-source spatiotemporal precipitation data. It addresses the technical challenges of existing technologies in dealing with the spatiotemporal dynamic evolution of precipitation, such as single data dimension, rigid claims thresholds, and lack of historical risk correlation. It achieves the technical effect of dynamically coupling multi-source heterogeneous precipitation data fusion analysis with historical claims spatial correlation, thereby significantly optimizing the efficiency of investigation resource allocation and the efficiency of insurance companies' disaster prevention and loss reduction actions while improving the accuracy of claims risk identification.

[0005] To achieve the above objectives, this invention provides a property insurance claims risk management method based on multi-source spatiotemporal precipitation data, comprising:

[0006] Obtain precipitation datasets collected by a multi-source precipitation monitoring system in the target area during the target time period;

[0007] Based on the continuous variation characteristics of the precipitation dataset in the temporal dimension, a precipitation temporal stability index is determined, and based on the discrete distribution characteristics of the precipitation dataset in the spatial dimension, a regional difference dispersion is constructed.

[0008] Based on the spatial overlay relationship between historical claims cases and the current precipitation grid, and combined with the intensity evolution pattern of the precipitation dataset, the spatial correlation of risk is obtained.

[0009] By integrating the precipitation time series stability index, the regional difference dispersion, and the risk spatial correlation, a dynamic correction coefficient for the claims threshold is generated.

[0010] The preset claim trigger threshold is dynamically adjusted based on the claim threshold dynamic correction coefficient to form a claim risk control management strategy.

[0011] Furthermore, when acquiring precipitation datasets collected by a multi-source precipitation monitoring system in the target area during the target time period, the following are included:

[0012] The multi-source precipitation monitoring system includes a satellite-based remote sensing observation platform, a ground-based rainfall sensor network, weather radar echo monitoring equipment, and a numerical weather prediction system.

[0013] The gridded inversion data of the satellite-based remote sensing observation platform, the station measured data of the ground-based rainfall sensing network, the quantitative estimation data of the weather radar echo monitoring equipment, and the model output data of the numerical weather prediction system are uniformly mapped to a standard spatiotemporal reference frame, which includes a fixed spatial resolution grid and equally spaced time slices.

[0014] The mapping multi-source data is subjected to consistency verification, and abnormal data sources that deviate from the preset deviation range of the overall distribution are removed to form the precipitation dataset.

[0015] Furthermore, when determining the precipitation time-series stability index based on the continuous variation characteristics of the precipitation dataset in the time-series dimension, the following steps are included:

[0016] Within the target time period, a forward time series analysis window is constructed with the current time as the endpoint, and the temporal gradient direction vector and temporal curvature change rate of precipitation at each grid point within the forward time series analysis window are calculated;

[0017] The directional coordination degree of the temporal gradient direction vectors of all grid points within the forward temporal analysis window is statistically analyzed, and combined with the second-order perturbation intensity of the temporal curvature change rate, the precipitation temporal stability index is generated by weighted fusion.

[0018] Furthermore, when constructing the regional disparity dispersion based on the spatial discrete characteristics of the precipitation dataset, the following steps are included:

[0019] The target area is divided into orthogonal grid cells along the latitude and longitude directions, and the average absolute deviation of precipitation at each grid point in the orthogonal grid cell relative to the median value within the cell is calculated.

[0020] Spatial discreteness assessment is performed on the average absolute deviation of all the orthogonal grid cells, and the assessment results are normalized and mapped to a preset standard interval to obtain the regional difference dispersion.

[0021] Furthermore, when obtaining the spatial correlation of risk based on the spatial overlay relationship between historical claims case areas and the current precipitation grid, combined with the intensity evolution pattern of the precipitation dataset, the following steps are included:

[0022] Extract spatial location information of claims events marked as rainfall-related damage from the historical claims case database to construct a set of historical risk locations;

[0023] Spatial matching is performed between the set of grid points in the current precipitation data that exceed the preset intensity threshold and the set of historical risk locations. The proportion of the number of matching points to the total number of historical risk locations is calculated as the location overlap rate.

[0024] The intensity center migration trajectory of the precipitation dataset is traced backward along the time axis. The spatiotemporal proximity of the intensity center migration trajectory with the historical risk location set is calculated. The location overlap rate and the spatiotemporal proximity are fused to generate the risk spatial correlation.

[0025] Furthermore, when spatially matching the set of grid points in the current precipitation data that exceed a preset intensity threshold with the set of historical risk locations, and calculating the proportion of the number of matching points to the total number of historical risk locations as the location overlap rate, the process includes:

[0026] Convert the spatial location information of each claim event in the historical risk location set into a geohash code to construct a historical risk location code library;

[0027] For each grid point in the set of grid points that exceed the preset intensity threshold at the current time, perform geohashing encoding conversion according to the same rules to obtain the grid point encoding set to be matched;

[0028] Calculate the coding similarity between the grid point coding set to be matched and the historical risk location coding library. Grid points with coding similarity higher than the similarity threshold are identified as duplicate qualified points. Calculate the proportion of the number of duplicate qualified points to the total number of the historical risk location set as the location overlap rate.

[0029] Furthermore, when tracing the intensity center migration trajectory of the precipitation dataset backward along the time axis and calculating the spatiotemporal proximity of the intensity center migration trajectory to the historical risk location set, the process includes:

[0030] Using the precipitation amount of each grid point in the precipitation dataset as a weighting factor, the intensity-weighted centroid coordinates at the current moment are calculated;

[0031] Multiple historical moments are traced back along the time axis with a fixed step size, and the intensity-weighted centroid coordinates of each historical moment are calculated to form an intensity center time series.

[0032] Vector trend extrapolation is performed on the intensity center time series to obtain the predicted migration direction vector. The spatial direction angle between the predicted migration direction vector and the location of each historical claim event is calculated. The weighted cumulative result of the cosine value of the spatial direction angle is used as the spatiotemporal proximity.

[0033] Furthermore, when generating the dynamic correction coefficient for the claims threshold by integrating the precipitation time-series stability index, the regional difference dispersion, and the risk spatial correlation, the following is included:

[0034] If the precipitation time series stability index is lower than the stability threshold and the regional difference dispersion is higher than the dispersion threshold, extreme precipitation pattern identification is triggered, and the dynamic correction coefficient of the claim threshold is determined according to the first power scaling relationship of the risk spatial correlation.

[0035] If the precipitation time series stability index is not lower than the stability threshold and the regional difference dispersion is not higher than the dispersion threshold, then conventional precipitation pattern identification is triggered, and the dynamic correction coefficient of the claim threshold is determined based on the weighted sum of the precipitation time series stability index and the regional difference dispersion.

[0036] Furthermore, when dynamically adjusting the preset claim trigger threshold based on the claim threshold dynamic correction coefficient to form a claim risk control management strategy, it includes:

[0037] Establish a piecewise linear mapping table between the dynamic correction coefficient of the claim threshold and the threshold adjustment step size. The piecewise linear mapping table contains multiple correction coefficient intervals and corresponding adjustment step size intervals.

[0038] Based on the correction coefficient range in which the claim threshold dynamic correction coefficient is located, the corresponding adjustment step range is queried, and the sum of the preset claim trigger threshold and the median value of the adjustment step range is used as the dynamically adjusted claim trigger threshold for output.

[0039] To achieve the above objectives, the present invention also provides a property insurance claims risk management system based on multi-source spatiotemporal precipitation data, comprising:

[0040] The data acquisition module is used to acquire precipitation datasets collected by the multi-source precipitation monitoring system in the target area during the target time period;

[0041] The precipitation distribution module is used to determine the precipitation temporal stability index based on the continuous change characteristics of the precipitation dataset in the temporal dimension, and to construct the regional difference dispersion based on the spatial distribution discrete characteristics of the precipitation dataset.

[0042] The spatial correlation module is used to obtain the spatial correlation degree of risk based on the spatial overlay relationship between the historical claims case area and the current precipitation grid, combined with the intensity evolution pattern of the precipitation dataset;

[0043] The comprehensive analysis module is used to integrate the precipitation time series stability index, the regional difference dispersion, and the risk spatial correlation to generate a dynamic correction coefficient for the claims threshold.

[0044] The claims management module is used to dynamically adjust the preset claims trigger threshold based on the claims threshold dynamic correction coefficient, thereby forming a claims risk control management strategy.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] This invention discloses a method and system for property insurance claims risk management based on multi-source spatiotemporal precipitation data. The method includes: acquiring a precipitation dataset of a target area within a target time period; determining a precipitation time-series stability index based on the continuous change characteristics of the precipitation dataset in the temporal dimension, and constructing a regional difference dispersion degree based on the spatial distribution discrete characteristics of the precipitation dataset; obtaining a risk spatial correlation degree based on historical claims case areas and the current precipitation grid, combined with the intensity evolution pattern of the precipitation dataset; generating a dynamic correction coefficient for claims thresholds by integrating the precipitation time-series stability index, regional difference dispersion degree, and risk spatial correlation degree; and dynamically adjusting a preset claims trigger threshold based on the dynamic correction coefficient for claims thresholds to form a claims risk control management strategy. This significantly optimizes the efficiency of investigation resource allocation and the efficiency of insurance companies' disaster prevention and mitigation actions while improving the accuracy of claims risk identification. Attached Figure Description

[0047] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0048] Figure 1 A flowchart illustrating the property insurance claims risk management method based on multi-source spatiotemporal precipitation data in an embodiment of the present invention is shown.

[0049] Figure 2 This diagram illustrates the structure of a property insurance claims risk management system based on multi-source spatiotemporal precipitation data in an embodiment of the present invention. Detailed Implementation

[0050] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0051] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0052] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0053] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0054] The following is a description of preferred embodiments of the present invention in conjunction with the accompanying drawings.

[0055] like Figure 1 As shown, embodiments of the present invention disclose a property insurance claims risk management method based on multi-source spatiotemporal precipitation data, including:

[0056] S110: Obtain the precipitation dataset collected by the multi-source precipitation monitoring system in the target area during the target time period;

[0057] S120: Based on the continuous change characteristics of the precipitation dataset in the temporal dimension, determine the precipitation temporal stability index; based on the spatial distribution discrete characteristics of the precipitation dataset, construct the regional difference dispersion.

[0058] S130: Based on the spatial overlay relationship between historical claims case areas and the current precipitation grid, and combined with the intensity evolution pattern of the precipitation dataset, obtain the spatial correlation of risk;

[0059] S140: By integrating the precipitation time series stability index, the regional difference dispersion, and the risk spatial correlation, a dynamic correction coefficient for the claims threshold is generated;

[0060] S150: Based on the dynamic correction coefficient of the claim threshold, the preset claim trigger threshold is dynamically adjusted to form a claim risk control management strategy.

[0061] In some embodiments of this application, the process of acquiring precipitation datasets collected by a multi-source precipitation monitoring system in a target area during a target time period includes:

[0062] The multi-source precipitation monitoring system includes a satellite-based remote sensing observation platform, a ground-based rainfall sensor network, weather radar echo monitoring equipment, and a numerical weather prediction system.

[0063] The gridded inversion data of the satellite-based remote sensing observation platform, the station measured data of the ground-based rainfall sensing network, the quantitative estimation data of the weather radar echo monitoring equipment, and the model output data of the numerical weather prediction system are uniformly mapped to a standard spatiotemporal reference frame, which includes a fixed spatial resolution grid and equally spaced time slices.

[0064] The mapping multi-source data is subjected to consistency verification, and abnormal data sources that deviate from the preset deviation range of the overall distribution are removed to form the precipitation dataset.

[0065] In this embodiment, the standard spatiotemporal reference framework is a grid with a spatial resolution of 3 kilometers and a temporal resolution of 1 hour. During unified mapping, bilinear interpolation is used to convert data of different resolutions to the standard grid. The fixed spatial resolution grid adopts the Lambert conformal projection coordinate system, with a grid size of 100×100. Equal-interval time slices are generated every hour starting from 00:00 each day, for a total of 24 slices per day. Consistency verification uses the spatial correlation coefficient method, calculating the correlation coefficient between each data source and the mean of multiple data sources. A data source with a correlation coefficient below 0.6 is considered an abnormal data source. After removing abnormal data sources, the arithmetic mean of the remaining data sources is used as the final precipitation value for that grid point at that time, forming a complete precipitation dataset.

[0066] The beneficial effects of the above technical solution are: by establishing a unified spatiotemporal benchmark framework and consistency verification mechanism, the spatial mismatch and temporal asynchrony of multi-source data are eliminated, the reliability and availability of precipitation datasets are improved, and a high-quality data foundation is provided for subsequent risk analysis.

[0067] In some embodiments of this application, when determining the precipitation time-series stability index based on the continuous variation characteristics of the precipitation dataset in the time-series dimension, the following steps are included:

[0068] Within the target time period, a forward time series analysis window is constructed with the current time as the endpoint, and the temporal gradient direction vector and temporal curvature change rate of precipitation at each grid point within the forward time series analysis window are calculated;

[0069] The directional coordination degree of the temporal gradient direction vectors of all grid points within the forward temporal analysis window is statistically analyzed, and combined with the second-order perturbation intensity of the temporal curvature change rate, the precipitation temporal stability index is generated by weighted fusion.

[0070] In this embodiment, the forward time series analysis window is set to the current time and the previous 6 hours, for a total of 7 time slices. The time series gradient direction vector is determined by comparing the direction of precipitation change between adjacent times; if precipitation increases, the direction vector value is 1; if it decreases, it is -1; and if it remains unchanged, it is 0. The time series curvature change rate is obtained by calculating the change rate of the gradient direction vector. The direction coherence is the consistency ratio of the gradient direction vectors of all grid points within the statistical window; for example, when more than 80% of the grid points have the same direction, the coherence value is 0.8. The second-order disturbance intensity is the absolute value of the average curvature change rate. The weighted fusion adopts a linear weighting method, with the direction coherence weight set to 0.6 and the second-order disturbance intensity weight set to 0.4. The final precipitation time series stability index ranges from 0 to 1.

[0071] The beneficial effects of the above technical solution are: by constructing a forward time series analysis window and quantifying the directional coordination degree and the intensity of second-order disturbance, a refined assessment of the stability of precipitation processes can be achieved, providing a quantitative basis for distinguishing between persistent rainstorms and localized showers, and improving the temporal resolution of risk identification.

[0072] In some embodiments of this application, when constructing the regional disparity dispersion based on the spatial discrete characteristics of the precipitation dataset, the following methods are included:

[0073] The target area is divided into orthogonal grid cells along the latitude and longitude directions, and the average absolute deviation of precipitation at each grid point in the orthogonal grid cell relative to the median value within the cell is calculated.

[0074] Spatial discreteness assessment is performed on the average absolute deviation of all the orthogonal grid cells, and the assessment results are normalized and mapped to a preset standard interval to obtain the regional difference dispersion.

[0075] In this embodiment, orthogonal grid cells are divided along the latitude and longitude directions. Each grid cell is set to a size of 0.05 degrees × 0.05 degrees, approximately 5 kilometers × 5 kilometers, and the target area is divided into 20 × 20 = 400 grid cells. The mean absolute deviation is calculated as the absolute difference between the precipitation at each grid point and the median value, and then the average of the absolute differences of all grid points within the cell is calculated. Spatial dispersion is assessed using the standard deviation method, and the standard deviation of the mean absolute deviation of the 400 grid cells is calculated. Normalization mapping linearly maps the standard deviation values ​​to the interval between 0 and 1.

[0076] The beneficial effects of the above technical solution are: by dividing the latitude and longitude orthogonal grid and calculating the dispersion of the median benchmark, the unevenness of the spatial distribution of precipitation can be effectively identified, overcoming the problem of insufficient sensitivity of the traditional mean benchmark to extreme precipitation grid points, and improving the ability to identify regional claims risks caused by local rainstorms.

[0077] In some embodiments of this application, when obtaining the spatial correlation of risk based on the spatial overlay relationship between historical claims case areas and the current precipitation grid, combined with the intensity evolution pattern of the precipitation dataset, the following steps are included:

[0078] Extract spatial location information of claims events marked as rainfall-related damage from the historical claims case database to construct a set of historical risk locations;

[0079] Spatial matching is performed between the set of grid points in the current precipitation data that exceed the preset intensity threshold and the set of historical risk locations. The proportion of the number of matching points to the total number of historical risk locations is calculated as the location overlap rate.

[0080] The intensity center migration trajectory of the precipitation dataset is traced backward along the time axis. The spatiotemporal proximity of the intensity center migration trajectory with the historical risk location set is calculated. The location overlap rate and the spatiotemporal proximity are fused to generate the risk spatial correlation.

[0081] In this embodiment, the historical claims database selects 1200 claims cases marked as rainstorm-related damage from the past five years, and extracts their latitude and longitude coordinates to construct a historical risk location set. When generating the risk spatial correlation degree by integrating the location overlap rate and spatiotemporal proximity: the location overlap rate is positively correlated and normalized to obtain a first mapping value, and the spatiotemporal proximity is negatively correlated and normalized to obtain a second mapping value. The weighted sum of the first and second mapping values ​​is calculated as the risk spatial correlation degree. For example, if the location overlap rate is 50%, the positive correlation normalization mapping uses a linear mapping method, linearly transforming the original overlap rate from 0% to 100% to the interval of 0 to 1. The transformation coefficient is determined based on the statistical analysis of the historical maximum overlap rate. When the overlap rate is 50%, the first mapping value is 0.5. If the spatiotemporal proximity is 0.75, the negative correlation normalization mapping uses an inverse linear mapping, setting the mapping coefficient to -0.8 according to the principle that the larger the proximity value, the smaller the mapping result. When the proximity is 0.75, the second mapping value is 0.4. The weight of the first mapping value is set to 0.6, and the weight of the second mapping value is set to 0.4. This weight is determined based on the relative importance of the location overlap rate to the spatial locking ability of claims risk and the spatiotemporal proximity to the ability to predict risk transmission. The risk spatial correlation degree is 0.46 obtained by weighted harmonic calculation.

[0082] The beneficial effects of the above technical solutions are: to achieve fault tolerance in spatial matching through geohashing encoding, to achieve forward-looking assessment of precipitation system movement through intensity center trajectory extrapolation, and to improve the ability to capture spatial transmission paths of precipitation-induced disaster risks by integrating static spatial correlation and dynamic temporal proximity through positive and negative correlation differential mapping and harmonization fusion mechanism.

[0083] In some embodiments of this application, when spatially matching the set of grid points in the current precipitation data that exceed a preset intensity threshold with the set of historical risk locations, and calculating the proportion of the number of matching duplicate points to the total number of historical risk locations as the location overlap rate, the following steps are included:

[0084] Convert the spatial location information of each claim event in the historical risk location set into a geohash code to construct a historical risk location code library;

[0085] For each grid point in the set of grid points that exceed the preset intensity threshold at the current time, perform geohashing encoding conversion according to the same rules to obtain the grid point encoding set to be matched;

[0086] Calculate the coding similarity between the grid point coding set to be matched and the historical risk location coding library. Grid points with coding similarity higher than the similarity threshold are identified as duplicate qualified points. Calculate the proportion of the number of duplicate qualified points to the total number of the historical risk location set as the location overlap rate.

[0087] In this embodiment, geohashing encoding uses 10-bit string precision. The historical risk location encoding library is generated by converting the coordinates of 1200 claim events, with each claim event corresponding to a unique encoding string. The preset intensity threshold is 30 mm, and the grid set includes all grid points with 24-hour precipitation exceeding 30 mm. The center coordinates of each grid point are converted into geohashing codes with the same 10-bit precision, forming a set of grid point codes to be matched. For example, grid coordinates (118.5, 35.2) are converted into the code "wx4g0r7xz0". The encoding similarity is calculated using Hamming distance, which is the number of different positions between two strings of characters, and the similarity threshold is set to 3 bits. The total number of historical risk location sets is 1200. If the number of overlapping qualified points is 600, the location overlap rate is 50%.

[0088] The beneficial effects of the above technical solution are: by using geo-hash encoding and Hamming distance similarity determination, it can achieve rapid spatial matching between historical claims locations and current precipitation grid points, improve matching efficiency and fault tolerance, and provide an efficient technical means for calculating location overlap rate.

[0089] In some embodiments of this application, when tracing the intensity center migration trajectory of the precipitation dataset backward along the time axis and calculating the spatiotemporal proximity of the intensity center migration trajectory to the historical risk location set, the method includes:

[0090] Using the precipitation amount of each grid point in the precipitation dataset as a weighting factor, the intensity-weighted centroid coordinates at the current moment are calculated;

[0091] Multiple historical moments are traced back along the time axis with a fixed step size, and the intensity-weighted centroid coordinates of each historical moment are calculated to form an intensity center time series.

[0092] Vector trend extrapolation is performed on the intensity center time series to obtain the predicted migration direction vector. The spatial direction angle between the predicted migration direction vector and the location of each historical claim event is calculated. The weighted cumulative result of the cosine value of the spatial direction angle is used as the spatiotemporal proximity.

[0093] In this embodiment, the weighting factor is normalized by dividing the precipitation at each grid point by the total precipitation at all grid points, with the weight value ranging from 0 to 1. When calculating the intensity-weighted centroid coordinates, the latitude and longitude of each grid point are multiplied by its weight and then summed to obtain the centroid coordinates. A fixed step size of 1 hour is set, and six historical moments are traced back to form a time series consisting of seven centroid coordinates. Vector trend extrapolation uses a straight line fitted with the centroid positions of the last three moments, and the direction of the line extension is used as the predicted migration direction vector. When calculating the spatial direction angle, the angle between the predicted migration vector and the vector pointing to each historical claim event is calculated; the cosine value is positive when the angle is less than 90 degrees and negative when it is greater than 90 degrees. The weighted cumulative weight is set as the probability of the occurrence of a historical claim event, and the probability value is determined by statistical analysis of the historical claim frequency at that location, ranging from 0 to 1.

[0094] The beneficial effects of the above technical solution are: by tracking the intensity-weighted centroid trajectory and extrapolating the vector trend, dynamic prediction of the movement of the precipitation intensity center can be achieved; by combining the cosine weighted accumulation of the direction angle, the spatiotemporal proximity of the predicted trajectory to the historical risk location can be quantified, thereby improving the ability to predict the spatial transmission of precipitation risk.

[0095] In some embodiments of this application, when generating a dynamic correction coefficient for the claims threshold by integrating the precipitation time-series stability index, the regional difference dispersion, and the risk spatial correlation, the following steps are included:

[0096] If the precipitation time series stability index is lower than the stability threshold and the regional difference dispersion is higher than the dispersion threshold, extreme precipitation pattern identification is triggered, and the dynamic correction coefficient of the claim threshold is determined according to the first power scaling relationship of the risk spatial correlation.

[0097] If the precipitation time series stability index is not lower than the stability threshold and the regional difference dispersion is not higher than the dispersion threshold, then conventional precipitation pattern identification is triggered, and the dynamic correction coefficient of the claim threshold is determined based on the weighted sum of the precipitation time series stability index and the regional difference dispersion.

[0098] In this embodiment, the stability threshold is set to 0.3, which is determined based on the distribution characteristics of the precipitation time-series stability index during historical heavy precipitation events. A stability index below 0.3 indicates that the precipitation process is extremely unstable. The dispersion threshold is set to 0.5, which is determined based on the mean plus one standard deviation of the regional difference dispersion in historical claims events. A dispersion above 0.5 indicates significant spatial differences in precipitation. The trigger condition for extreme precipitation pattern identification is that both conditions are met simultaneously, i.e., the precipitation is both unstable and has large spatial differences. The first power scaling relationship uses square root scaling, i.e., the correction coefficient is equal to the square root of the risk spatial correlation. This scaling method ensures that the correction coefficient is not excessively amplified in extreme modes, and the value range is limited to between 0.5 and 1.2. The trigger condition for normal precipitation pattern identification is that both conditions are not met simultaneously, i.e., the precipitation is relatively stable and the spatial differences are moderate. The weight of the precipitation time-series stability index is set to 0.4, and the weight of the regional difference dispersion is set to 0.6. The weight settings are determined based on the historical statistical analysis of the impact of these two parameters on claims risk.

[0099] The beneficial effects of the above technical solution are: automatic identification and classification of precipitation patterns are achieved through dual threshold determination; power scaling is used to control the correction amplitude for extreme precipitation patterns; and weighted balancing of the influence of multiple parameters is used for conventional patterns, thereby improving the adaptability and rationality of the correction coefficient under different precipitation intensities.

[0100] In some embodiments of this application, when dynamically adjusting the preset claim trigger threshold based on the claim threshold dynamic correction coefficient to form a claim risk control management strategy, the following is included:

[0101] Establish a piecewise linear mapping table between the dynamic correction coefficient of the claim threshold and the threshold adjustment step size. The piecewise linear mapping table contains multiple correction coefficient intervals and corresponding adjustment step size intervals.

[0102] Based on the correction coefficient range in which the claim threshold dynamic correction coefficient is located, the corresponding adjustment step range is queried, and the sum of the preset claim trigger threshold and the median value of the adjustment step range is used as the dynamically adjusted claim trigger threshold for output.

[0103] In this embodiment, the piecewise linear mapping table is divided into 5 intervals: a correction coefficient of 0.5 to 0.7 corresponds to an adjustment step of -15 to -10 mm, with a midpoint of -12.5 mm; a correction coefficient of 0.7 to 0.9 corresponds to an adjustment step of -10 to -5 mm, with a midpoint of -7.5 mm; a correction coefficient of 0.9 to 1.1 corresponds to an adjustment step of -5 to 5 mm, with a midpoint of 0 mm; a correction coefficient of 1.1 to 1.3 corresponds to an adjustment step of 5 to 10 mm, with a midpoint of 7.5 mm; and a correction coefficient of 1.3 to 1.5 corresponds to an adjustment step of 10 to 15 mm, with a midpoint of 12.5 mm. A negative step indicates a lower threshold, and a positive step indicates an upper threshold. For example, if the initial value of the preset claim trigger threshold (precipitation value) is 50 mm, when the correction coefficient is 0.8, it falls within the 0.7 to 0.9 interval, with a midpoint of the adjustment step of -7.5 mm. The dynamically adjusted claim trigger threshold is 42.5 mm. This segmented mapping table is determined based on historical statistics of claim hit rates under different correction coefficients, ensuring that the adjusted threshold matches the risk level. After the threshold adjustment is output, the claims decision-making system uses the new threshold as the effective threshold for subsequent grid point comparisons. As mentioned above, the dynamically adjusted claim trigger threshold is 42.5 mm, and the decision-making system sets this value as the effective threshold. Hourly scanning revealed that 12 grid points in a certain county had rainfall exceeding 42.5 mm, and the system automatically marked this area as a risk warning zone. A reminder SMS containing risk avoidance guidelines was sent to insured customers in this area, instructing them to carry out precise disaster prevention and loss reduction actions (such as moving cars in underground garages and transferring supplies).

[0104] The beneficial effects of the above technical solution are: by using a piecewise linear mapping table to achieve standardized conversion of the correction coefficient to the threshold adjustment amount, the threshold adjustment process can be quantified and traced, avoiding subjective judgment, improving the objectivity and consistency of threshold adjustment, realizing fully automated response from risk monitoring to early warning notification, investigation and dispatch, case acceptance and fund transfer, shortening early warning handling time, and improving the timeliness of claims service and customer satisfaction.

[0105] To further illustrate the technical concept of this invention, the technical solution of this invention will now be described in conjunction with specific application scenarios.

[0106] Correspondingly, such as Figure 2 As shown, this application also provides a property insurance claims risk management system based on multi-source spatiotemporal precipitation data, including:

[0107] The data acquisition module is used to acquire precipitation datasets collected by the multi-source precipitation monitoring system in the target area during the target time period;

[0108] The precipitation distribution module is used to determine the precipitation temporal stability index based on the continuous change characteristics of the precipitation dataset in the temporal dimension, and to construct the regional difference dispersion based on the spatial distribution discrete characteristics of the precipitation dataset.

[0109] The spatial correlation module is used to obtain the spatial correlation degree of risk based on the spatial overlay relationship between the historical claims case area and the current precipitation grid, combined with the intensity evolution pattern of the precipitation dataset;

[0110] The comprehensive analysis module is used to integrate the precipitation time series stability index, the regional difference dispersion, and the risk spatial correlation to generate a dynamic correction coefficient for the claims threshold.

[0111] The claims management module is used to dynamically adjust the preset claims trigger threshold based on the claims threshold dynamic correction coefficient, thereby forming a claims risk control management strategy.

[0112] In the description of the above embodiments, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0113] Although the invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the embodiments disclosed in this invention can be combined with each other in any way. The fact that not all of these combinations are described in this specification is merely for the sake of brevity and resource conservation.

[0114] It will be understood by those skilled in the art that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A property insurance claims risk management method based on multi-source spatiotemporal precipitation data, characterized in that, include: Obtain precipitation datasets collected by a multi-source precipitation monitoring system in the target area during the target time period; Based on the continuous variation characteristics of the precipitation dataset in the temporal dimension, a precipitation temporal stability index is determined, and based on the discrete distribution characteristics of the precipitation dataset in the spatial dimension, a regional difference dispersion is constructed. Based on the spatial overlay relationship between historical claims cases and the current precipitation grid, and combined with the intensity evolution pattern of the precipitation dataset, the spatial correlation of risk is obtained. By integrating the precipitation time series stability index, the regional difference dispersion, and the risk spatial correlation, a dynamic correction coefficient for the claims threshold is generated. The preset claim trigger threshold is dynamically adjusted based on the claim threshold dynamic correction coefficient to form a claim risk control management strategy.

2. The property insurance claims risk management method based on multi-source spatiotemporal precipitation data according to claim 1, characterized in that, When acquiring precipitation datasets collected by a multi-source precipitation monitoring system in the target area during the target time period, the following should be included: The multi-source precipitation monitoring system includes a satellite-based remote sensing observation platform, a ground-based rainfall sensor network, weather radar echo monitoring equipment, and a numerical weather prediction system. The gridded inversion data of the satellite-based remote sensing observation platform, the station measured data of the ground-based rainfall sensing network, the quantitative estimation data of the weather radar echo monitoring equipment, and the model output data of the numerical weather prediction system are uniformly mapped to a standard spatiotemporal reference frame, which includes a fixed spatial resolution grid and equally spaced time slices. The mapping multi-source data is subjected to consistency verification, and abnormal data sources that deviate from the preset deviation range of the overall distribution are removed to form the precipitation dataset.

3. The property insurance claims risk management method based on multi-source spatiotemporal precipitation data according to claim 1, characterized in that, When determining the precipitation time-series stability index based on the continuous variation characteristics of the precipitation dataset in the time-series dimension, the following steps are included: Within the target time period, a forward time series analysis window is constructed with the current time as the endpoint, and the temporal gradient direction vector and temporal curvature change rate of precipitation at each grid point within the forward time series analysis window are calculated; The directional coordination degree of the temporal gradient direction vectors of all grid points within the forward temporal analysis window is statistically analyzed, and combined with the second-order perturbation intensity of the temporal curvature change rate, the precipitation temporal stability index is generated by weighted fusion.

4. The property insurance claims risk management method based on multi-source spatiotemporal precipitation data according to claim 1, characterized in that, When constructing the regional disparity dispersion based on the spatial distribution discrete characteristics of the precipitation dataset, the following steps are included: The target area is divided into orthogonal grid cells along the latitude and longitude directions, and the average absolute deviation of precipitation at each grid point in the orthogonal grid cell relative to the median value within the cell is calculated. Spatial discreteness assessment is performed on the average absolute deviation of all the orthogonal grid cells, and the assessment results are normalized and mapped to a preset standard interval to obtain the regional difference dispersion.

5. The property insurance claims risk management method based on multi-source spatiotemporal precipitation data according to claim 1, characterized in that, When obtaining the spatial correlation of risk based on the spatial overlay relationship between historical claims case areas and the current precipitation grid, combined with the intensity evolution pattern of the precipitation dataset, the following are included: Extract spatial location information of claims events marked as rainfall-related damage from the historical claims case database to construct a set of historical risk locations; Spatial matching is performed between the set of grid points in the current precipitation data that exceed the preset intensity threshold and the set of historical risk locations. The proportion of the number of matching points to the total number of historical risk locations is calculated as the location overlap rate. The intensity center migration trajectory of the precipitation dataset is traced backward along the time axis. The spatiotemporal proximity of the intensity center migration trajectory with the historical risk location set is calculated. The location overlap rate and the spatiotemporal proximity are fused to generate the risk spatial correlation.

6. The property insurance claims risk management method based on multi-source spatiotemporal precipitation data according to claim 5, characterized in that, When spatially matching the set of grid points in the current precipitation dataset that exceed a preset intensity threshold with the set of historical risk locations, and calculating the proportion of the number of matching points to the total number of historical risk locations as the location overlap rate, the following steps are included: Convert the spatial location information of each claim event in the historical risk location set into a geohash code to construct a historical risk location code library; For each grid point in the set of grid points that exceed the preset intensity threshold at the current time, perform geohashing encoding conversion according to the same rules to obtain the grid point encoding set to be matched; Calculate the coding similarity between the grid point coding set to be matched and the historical risk location coding library. Grid points with coding similarity higher than the similarity threshold are identified as duplicate qualified points. Calculate the proportion of the number of duplicate qualified points to the total number of the historical risk location set as the location overlap rate.

7. The property insurance claims risk management method based on multi-source spatiotemporal precipitation data according to claim 6, characterized in that, When tracing the intensity center migration trajectory of the precipitation dataset backward along the time axis and calculating the spatiotemporal proximity of the intensity center migration trajectory to the historical risk location set, the following steps are included: Using the precipitation amount of each grid point in the precipitation dataset as a weighting factor, the intensity-weighted centroid coordinates at the current moment are calculated; Multiple historical moments are traced back along the time axis with a fixed step size, and the intensity-weighted centroid coordinates of each historical moment are calculated to form an intensity center time series. Vector trend extrapolation is performed on the intensity center time series to obtain the predicted migration direction vector. The spatial direction angle between the predicted migration direction vector and the location of each historical claim event is calculated. The weighted cumulative result of the cosine value of the spatial direction angle is used as the spatiotemporal proximity.

8. The property insurance claims risk management method based on multi-source spatiotemporal precipitation data according to claim 1, characterized in that, When generating a dynamic correction coefficient for the claims threshold by integrating the precipitation time-series stability index, the regional difference dispersion, and the risk spatial correlation, the following are included: If the precipitation time series stability index is lower than the stability threshold and the regional difference dispersion is higher than the dispersion threshold, extreme precipitation pattern identification is triggered, and the dynamic correction coefficient of the claim threshold is determined according to the first power scaling relationship of the risk spatial correlation. If the precipitation time series stability index is not lower than the stability threshold and the regional difference dispersion is not higher than the dispersion threshold, then conventional precipitation pattern identification is triggered, and the dynamic correction coefficient of the claim threshold is determined based on the weighted sum of the precipitation time series stability index and the regional difference dispersion.

9. The property insurance claims risk management method based on multi-source spatiotemporal precipitation data according to claim 1, characterized in that, When dynamically adjusting the preset claim trigger threshold based on the claim threshold dynamic correction coefficient to form a claim risk control management strategy, the strategy includes: Establish a piecewise linear mapping table between the dynamic correction coefficient of the claim threshold and the threshold adjustment step size. The piecewise linear mapping table contains multiple correction coefficient intervals and corresponding adjustment step size intervals. Based on the correction coefficient range in which the claim threshold dynamic correction coefficient is located, the corresponding adjustment step range is queried, and the sum of the preset claim trigger threshold and the median value of the adjustment step range is used as the dynamically adjusted claim trigger threshold for output.

10. A property insurance claims risk management system based on multi-source spatiotemporal precipitation data, applied to the property insurance claims risk management method based on multi-source spatiotemporal precipitation data as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire precipitation datasets collected by the multi-source precipitation monitoring system in the target area during the target time period; The precipitation distribution module is used to determine the precipitation temporal stability index based on the continuous change characteristics of the precipitation dataset in the temporal dimension, and to construct the regional difference dispersion based on the spatial distribution discrete characteristics of the precipitation dataset. The spatial correlation module is used to obtain the spatial correlation degree of risk based on the spatial overlay relationship between the historical claims case area and the current precipitation grid, combined with the intensity evolution pattern of the precipitation dataset; The comprehensive analysis module is used to integrate the precipitation time series stability index, the regional difference dispersion, and the risk spatial correlation to generate a dynamic correction coefficient for the claims threshold. The claims management module is used to dynamically adjust the preset claims trigger threshold based on the claims threshold dynamic correction coefficient, thereby forming a claims risk control management strategy.

Images