Insurance premium rate self-adaptive calculation method and system based on dynamic bayesian network

Abstract

This invention discloses an adaptive insurance premium calculation method and system based on a dynamic Bayesian network, belonging to the field of insurance premium optimization technology. The method includes: dynamically collecting multi-source heterogeneous information through a multi-source acquisition interface and constructing a dynamic Bayesian network based on a preset network structure; obtaining real-time risk assessment results through the dynamic Bayesian network and processing them to obtain a real-time risk vector; evaluating and analyzing the real-time risk vector using a historical risk database to obtain a real-time risk adjustment coefficient; and adjusting a predetermined insurance premium to obtain a dynamic insurance premium. This invention solves the technical problems in existing technologies where insurance premium calculation relies on fixed rules, making it difficult to integrate multi-source information and respond to risk changes in real time. It achieves the function of integrating multi-source information based on a dynamic Bayesian network, assessing risks in real time, and dynamically adjusting premiums, thus improving the accuracy and timeliness of insurance premium optimization.

Description

Technical Field

[0001] This invention relates to the field of insurance premium optimization technology, specifically to an adaptive insurance premium calculation method and system based on dynamic Bayesian networks. Background Technology

[0002] In the insurance industry, the calculation of insurance premiums needs to take into account a variety of risk factors. Traditional pricing methods often rely on fixed rate tables and static risk assessment models, which are difficult to respond to dynamic risk factors such as changes in user behavior, adjustments in disease trends, and updates in policies and regulations in real time. This results in discrepancies between insurance premiums and the actual risk environment, insufficient accuracy and flexibility in risk prediction, and an inability to achieve adaptive optimization of insurance pricing, making it difficult to meet the insurance pricing needs under the current and future risk environments.

[0003] Existing technologies suffer from technical problems such as relying on fixed rules for insurance premium calculation, difficulty in integrating multi-source information, and inability to respond to changes in risk in real time. Summary of the Invention

[0004] This application provides an adaptive calculation method and system for insurance rates based on dynamic Bayesian networks, which addresses the technical problems in existing technologies where insurance rate calculations rely on fixed rules, are difficult to integrate multi-source information, and cannot respond to risk changes in real time.

[0005] In view of the above problems, this application provides an adaptive calculation method and system for insurance rates based on dynamic Bayesian networks.

[0006] The first aspect of this application provides an adaptive calculation method for insurance premium rates based on dynamic Bayesian networks, the method comprising:

[0007] Multi-source heterogeneous information is dynamically collected through a multi-source acquisition interface, and a dynamic Bayesian network is constructed in conjunction with a preset network structure. Real-time risk assessment results are obtained through the dynamic Bayesian network, and a risk vectorization strategy is invoked to process the real-time risk assessment results to obtain a real-time risk vector. The real-time risk vector is evaluated and analyzed in conjunction with a historical risk database to obtain a real-time risk adjustment coefficient. The predetermined insurance premium rate is adjusted based on the real-time risk adjustment coefficient to obtain a dynamic insurance premium rate.

[0008] A second aspect of this application provides an adaptive insurance premium calculation system based on dynamic Bayesian networks, the system comprising:

[0009] The system includes a dynamic Bayesian network construction module, which dynamically collects heterogeneous information from multiple sources through a multi-source acquisition interface and constructs a dynamic Bayesian network based on a preset network structure; a real-time risk vector acquisition module, which obtains real-time risk assessment results through the dynamic Bayesian network and processes these results using a risk vectorization strategy to obtain a real-time risk vector; a risk adjustment coefficient acquisition module, which evaluates and analyzes the real-time risk vector using a historical risk database to obtain a real-time risk adjustment coefficient; and a dynamic insurance premium rate acquisition module, which adjusts a predetermined insurance premium rate based on the real-time risk adjustment coefficient to obtain a dynamic insurance premium rate.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] Multi-source heterogeneous information is dynamically collected through a multi-source acquisition interface, and a dynamic Bayesian network is constructed based on a preset network structure. Real-time risk assessment results are obtained through this dynamic Bayesian network and processed to obtain a real-time risk vector. This real-time risk vector is then evaluated and analyzed using a historical risk database to obtain a real-time risk adjustment coefficient. Finally, the predetermined insurance premium rate is adjusted to obtain a dynamic insurance premium rate. This achieves the technical effect of integrating multi-source information based on a dynamic Bayesian network, assessing risk in real time, and dynamically adjusting rates, thus improving the accuracy and timeliness of insurance premium optimization. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating the adaptive insurance rate calculation method based on dynamic Bayesian networks provided in this application embodiment;

[0014] Figure 2 This is a schematic diagram of the structure of the adaptive insurance rate calculation system based on dynamic Bayesian network provided in the embodiments of this application.

[0015] Figure labeling: Dynamic Bayesian network construction module 10, real-time risk vector acquisition module 20, risk adjustment coefficient acquisition module 30, dynamic insurance premium rate acquisition module 40. Detailed Implementation

[0016] This application provides an adaptive calculation method and system for insurance rates based on dynamic Bayesian networks, which addresses the technical problems in existing technologies where insurance rate calculations rely on fixed rules, are difficult to integrate multi-source information, and cannot respond to risk changes in real time.

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0018] Example 1, as Figure 1 As shown, this application provides an adaptive insurance premium calculation method based on dynamic Bayesian networks, the method comprising:

[0019] Step S100: Dynamically collect multi-source heterogeneous information through the multi-source acquisition interface, and construct a dynamic Bayesian network by combining it with a preset network structure.

[0020] Specifically, user behavior information, disease prevalence information, and policy change information are dynamically collected through the user behavior monitoring interface, public health database interface, and government portal website interface in the multi-source acquisition interface, respectively, to construct multi-source heterogeneous information. Then, a standardized processing strategy including time alignment, missing value imputation, and text structuring transformation is invoked to perform standardized preprocessing on the multi-source heterogeneous information to obtain the target multi-source information, which is used as the input information for the observation layer nodes in the preset network structure. The preset network structure includes hidden layer nodes storing a predetermined potential risk database and intent layer nodes storing a predetermined insurance product database. A dynamic Bayesian network is constructed based on the observation layer nodes, hidden layer nodes, and intent layer nodes.

[0021] Step S200: Obtain the real-time risk assessment result through the dynamic Bayesian network, and call the risk vectorization strategy to process the real-time risk assessment result to obtain the real-time risk vector.

[0022] Specifically, a first predetermined risk is extracted from predetermined potential risk types (including at least user behavior risk, disease infection risk, and policy and regulatory risk) through a dynamic Bayesian network. A first risk factor set for the first predetermined risk is matched in a predetermined potential risk database. Based on the first risk factor set, the target multi-source information is traversed and filtered to obtain a first factor parameter set. The first factor parameter set is analyzed to obtain a first real-time risk index for the first predetermined risk. It is then determined whether the first real-time risk index reaches a predetermined index limit stored in the intent layer node. If it does, the first mapping relationship between the first predetermined risk and the first real-time risk index is added to the real-time risk assessment result. Subsequently, a risk vectorization strategy is invoked, which contains a risk type order. Based on this order, the first mapping relationship is sorted to obtain a real-time risk vector.

[0023] Step S300: Combine the historical risk database to evaluate and analyze the real-time risk vector to obtain the real-time risk adjustment coefficient.

[0024] Specifically, the historical risk vector that is most similar to the real-time risk vector is searched through the historical risk database. The dynamic time-normalized distance between the historical risk vector and the real-time risk vector is calculated and normalized to obtain the real-time risk confidence level. This real-time risk confidence level is used as the real-time risk adjustment coefficient. At the same time, the historical insurance products corresponding to the historical risk vector are matched, and the real-time insurance premium rate of the historical insurance product matched in the predetermined insurance product database is determined as the predetermined insurance premium rate.

[0025] Step S400: Adjust the predetermined insurance premium rate based on the real-time risk adjustment coefficient to obtain the dynamic insurance premium rate.

[0026] Specifically, based on the obtained real-time risk adjustment coefficient (i.e., real-time risk confidence level), the predetermined insurance premium rate (i.e., the real-time insurance premium rate of the historical insurance product corresponding to the historical risk vector in the predetermined insurance product database) is adjusted. By linking the predetermined insurance premium rate with the real-time risk adjustment coefficient (by multiplying proportionally, etc.), a dynamic insurance premium rate that can dynamically adapt to the current risk situation is finally obtained, realizing the adaptive optimization of the insurance premium rate based on the real-time risk assessment results.

[0027] In one possible implementation, step S100 further includes:

[0028] Step S110: Dynamically collect user behavior information through the user behavior monitoring interface in the multi-source acquisition interface.

[0029] Step S120: Disease epidemic information is dynamically collected through the public health database interface in the multi-source acquisition interface.

[0030] Step S130: Dynamically collect policy change information through the government portal website interface in the multi-source collection interface.

[0031] Step S140: Based on the user behavior information, the disease prevalence information, and the policy change information, construct the multi-source heterogeneous information.

[0032] Specifically, relying on the user behavior monitoring interface integrated in the multi-source acquisition interface, and by connecting to the operation log system of the user's insurance terminal (such as APP, web terminal), the interaction record module of the claims business processing platform, and the health data interface of wearable devices, the system can capture dynamic information such as the user's insurance operation trajectory, claims application process nodes, and health monitoring data (such as exercise frequency and rest records) in real time. After interface protocol conversion and data format encapsulation, a structured user behavior information dataset is formed.

[0033] By leveraging the public health database interface integrated in the multi-source acquisition interface, and by connecting to official data sources such as national or local CDC databases and public health monitoring platforms via API interfaces, data on the incidence rate, distribution of epidemic areas, transmission trends, and epidemic early warning information of various diseases can be captured in real time. After data format conversion and field mapping processing by the interface, a standardized disease epidemic information dataset is formed, enabling the continuous collection of dynamic information related to disease epidemics.

[0034] Relying on the government portal website interface integrated in the multi-source data collection interface, and by connecting to the information release section API of the official websites of national and local government departments (such as the State Financial Supervision and Administration Bureau and the Social Insurance Administration Bureau), it can capture policy and regulation updates, regulatory revisions, and rate adjustment guidelines related to the insurance industry in real time. After interface protocol parsing and text information extraction processing, the unstructured policy text is converted into a structured policy change information dataset containing fields such as effective date, adjustment clauses, and scope of application, thus achieving continuous collection of policy dynamics.

[0035] User behavior information collected through the user behavior monitoring interface, disease epidemic information obtained through the public health database interface, and policy change information obtained through the government portal website interface are aggregated. These information sources are different and their data types are different (e.g., user behavior information is mostly structured operation records, disease epidemic information includes statistical data, and policy change information is mostly text content). Through a data aggregation mechanism, the three types of information are integrated into a unified dataset, forming multi-source heterogeneous information containing multi-source and multi-type data, providing the original data foundation for subsequent standardized processing.

[0036] In one possible implementation, step S140 further includes:

[0037] Step S141: Retrieve the standardization processing strategy to perform standardization preprocessing on the multi-source heterogeneous information to obtain target multi-source information, and use the target multi-source information as input information for the observation layer nodes in the preset network structure.

[0038] Step S142: Wherein, the preset network structure includes a hidden layer node storing a predetermined potential risk database and an intent layer node storing a predetermined insurance product database.

[0039] Step S143: Construct the dynamic Bayesian network based on the observation layer node, the hidden layer node, and the intent layer node.

[0040] Specifically, a standardized processing strategy, including time alignment, missing value imputation, and text structuring transformation, is invoked to perform standardized preprocessing on multi-source heterogeneous information composed of user behavior information, disease epidemic information, and policy change information. Time alignment is used to unify the time dimension of data from different sources, and missing value imputation is used to handle the missing parts in the data. Unstructured textual information (such as policy change text) is converted into structured data. After the above processing, the target multi-source information is obtained, and this target multi-source information is used as the input information of the observation layer nodes in the preset network structure, providing standardized basic data for the construction of dynamic Bayesian networks.

[0041] The aforementioned preset network structure comprises nodes at two key levels: first, a hidden layer node, which stores a database of predetermined potential risks, covering various types of predetermined potential risks such as user behavior risks, disease infection risks, and policy and regulatory risks, as well as related data on their corresponding risk factors; and second, an intent layer node, which stores a database of predetermined insurance products, containing basic information, premium rates, and terms and conditions for various insurance products. These two layers of nodes, together with the observation layer node that receives multi-source information from the target, constitute the core components of the preset network structure.

[0042] Using the observation layer nodes that receive multi-source target information as the data input basis, the observation layer nodes are associated with the hidden layer nodes that store a predetermined potential risk database. This allows the standardized information from the observation layer to provide a basis for the analysis of potential risks in the hidden layer nodes. At the same time, the hidden layer nodes are connected with the intent layer nodes that store a predetermined insurance product database, allowing potential risk data to form a mapping relationship with insurance product information. By defining the probabilistic dependencies and state transition rules between these three layers of nodes, a dynamic Bayesian network comprising the observation layer, hidden layer, and intent layer is constructed to achieve the fusion processing of multi-source information and support for risk assessment.

[0043] In one possible implementation, step S141 further includes:

[0044] The standardized processing strategies include time alignment, missing value imputation, and text structuring transformation.

[0045] Specifically, the standardized processing strategy includes three key steps: time alignment, which involves adjusting multi-source heterogeneous information from different channels such as user behavior monitoring, public health databases, and government portals to a consistent time dimension by unifying timestamp formats and calibrating data collection cycles, thereby eliminating the impact of time deviations on subsequent analysis; missing value imputation, which involves supplementing data records with mean imputation, interpolation, or inference methods based on similar data patterns to avoid analytical errors caused by missing values, addressing potential data gaps in multi-source heterogeneous information; and text structuring transformation, which involves converting unstructured text data such as policy change information into structured data containing fields such as policy type, adjustment content, and effective date through keyword extraction, entity recognition, and semantic parsing, enabling all types of information to participate in subsequent dynamic Bayesian network operations in a standardized format.

[0046] In one possible implementation, step S200 further includes:

[0047] Step S210: Extract the first predetermined risk from the predetermined potential risk types.

[0048] Step S220: Match the first risk factor set of the first predetermined risk in the predetermined potential risk database.

[0049] Step S230: Based on the first risk factor set, traverse and filter the target multi-source information to obtain the first factor parameter set.

[0050] Step S240: Analyze the first factor parameter set to obtain the first real-time risk index of the first predetermined risk.

[0051] Step S250: Determine whether the first real-time risk index has reached a predetermined index limit, wherein the predetermined index limit is stored in the intent layer node.

[0052] Step S260: If the first real-time risk index reaches the predetermined index limit, then the first mapping relationship between the first predetermined risk and the first real-time risk index is added to the real-time risk assessment result.

[0053] Specifically, one type of risk is selected from the predetermined potential risk types as the first predetermined risk. The predetermined potential risk types at least cover user behavior risks, disease infection risks, and policy and regulatory risks. By clearly extracting such a specific risk type, the foundation is laid for further analysis and processing of the risk.

[0054] Based on the extracted first predetermined risk (such as user behavior risk), a database retrieval algorithm is used to perform a targeted query on the predetermined potential risk database stored in the hidden layer node. The preset risk-factor mapping rules are invoked to match the set of various influencing factors directly related to the first predetermined risk, namely the first risk factor set (for example, if the first predetermined risk is user behavior risk, its first risk factor set may include user insurance frequency, claims history, health management behavior records, etc.). This set is then output in the form of structured data to provide a basis for the subsequent screening of target multi-source information.

[0055] Using the obtained first risk factor set as the screening basis, a comprehensive traversal of the target multi-source information (including standardized data such as user behavior, disease prevalence, and policy changes) obtained after standardized preprocessing is performed to extract specific parameters directly related to each factor in the first risk factor set. For example, if the first risk factor set involves factors related to user health behavior, parameters such as physical examination frequency and exercise data are screened from the user behavior information of the target multi-source information. Finally, these parameters are integrated to form the first factor parameter set corresponding to the first predetermined risk, providing specific data support for the subsequent analysis of the real-time risk index.

[0056] For the obtained first factor parameter set, the risk assessment algorithm preset in the dynamic Bayesian network is used to perform quantitative analysis and weighted calculation on each specific parameter in the parameter set (such as user behavior data, disease epidemic data, policy-related data, etc. related to the first predetermined risk). By multiplying each parameter with the corresponding risk impact weight and summing them up, a quantitative indicator that can reflect the current real-time status of the first predetermined risk is obtained, namely the first real-time risk index. This index can intuitively reflect the current severity or probability of occurrence of the risk.

[0057] The data comparison algorithm is invoked to compare the obtained first real-time risk index with the predetermined index limit stored in the intent layer node. Through the set comparison logic (such as greater than or equal to, equal to, etc.), it is determined whether the first real-time risk index has reached the predetermined index limit. The predetermined index limit is a threshold data that is pre-set and stored in the intent layer node based on factors such as the characteristics of the insurance product and the risk tolerance range.

[0058] When the first real-time risk index is determined to reach the predetermined index limit stored in the intent layer node, the mapping relationship generation mechanism is triggered. The identification information of the first predetermined risk (such as the user behavior risk code) is associated and bound with the corresponding first real-time risk index value to form a first mapping relationship data structure containing the risk type and risk level. Then, the mapping relationship is added to the dataset of the real-time risk assessment results through the data writing interface to ensure that the real-time risk assessment results completely record the risks that reach the threshold and their quantitative indicators.

[0059] In one possible implementation, step S210 further includes:

[0060] Step S211: The predetermined potential risk types include at least user behavior risk, disease infection risk, and policy and regulatory risk.

[0061] Specifically, the predetermined potential risk types are the basic classifications for risk assessment using dynamic Bayesian networks, which cover at least three core risk categories: user behavior risk, mainly related to the risks arising from various user behaviors in insurance-related activities (such as insurance application, claims, health management, etc.); disease infection risk, involving the impact of the epidemic trends and infection probabilities of various diseases on insurance; and policy and regulatory risk, which is the risk that may be caused by policy changes and regulatory revisions related to the insurance industry. These risk types together constitute the basic scope of predetermined potential risks, providing a clear direction for subsequent risk analysis.

[0062] In one possible implementation, step S200 further includes:

[0063] Step S270: The risk vectorization strategy contains a risk type order, and the first mapping relationship is sorted based on the risk type order to obtain the real-time risk vector.

[0064] Specifically, the risk vectorization strategy pre-stores a fixed order of risk types (such as user behavior risk, disease infection risk, and policy and regulatory risk). After obtaining the first mapping relationship in the real-time risk assessment results, these mapping relationships are sorted according to the order of risk types, so that the mapping relationships of different types of risks are arranged in a preset order, and finally an ordered real-time risk vector is formed. This vector can clearly and systematically reflect the real-time status of various risks.

[0065] In one possible implementation, step S300 further includes:

[0066] Step S310: Traverse the historical risk database to obtain the historical risk vector that is most similar to the real-time risk vector.

[0067] Step S320: Calculate the dynamic time-normalized distance between the historical risk vector and the real-time risk vector, and normalize it to obtain the real-time risk confidence.

[0068] Step S330: Use the real-time risk confidence level as the real-time risk adjustment coefficient.

[0069] Specifically, all historical risk vectors stored in the historical risk database are accessed and extracted one by one. Each historical risk vector is compared with the current real-time risk vector in terms of features. The degree of closeness is judged by calculating the similarity measure between the two in the vector space (such as cosine similarity, Euclidean distance, etc.). After a comprehensive traversal and comparison, the historical risk vector that is closest to the real-time risk vector in terms of feature distribution and has the highest similarity is selected.

[0070] For the most similar historical risk vector and the current real-time risk vector, the distance between them is calculated using a dynamic time warping algorithm. This algorithm measures the similarity between vector sequences of different lengths or time distributions by finding the optimal time alignment path. After obtaining the dynamic time warping distance, the distance value is transformed to the [0, 1] interval using the Min-Max normalization method to generate the real-time risk confidence score. The smaller the distance, the higher the normalized confidence score, thereby quantifying the similarity between real-time risk and historical risk.

[0071] After obtaining the real-time risk confidence level, the confidence level value is directly determined as the real-time risk adjustment coefficient without any additional conversion or calculation process. This allows the confidence level, which reflects the similarity between historical risk and real-time risk, to be directly used for subsequent adjustments to the predetermined insurance premium rate, becoming a key parameter connecting risk assessment results and premium calculation.

[0072] In one possible implementation, step S400 further includes:

[0073] Step S410: Match the historical insurance products corresponding to the historical risk vector, and use the real-time insurance premium rate of the historical insurance products matched in the predetermined insurance product database as the predetermined insurance premium rate.

[0074] Specifically, the corresponding historical insurance product record is retrieved by querying the historical risk database using the unique identifier of the historical risk vector (such as vector ID), and the feature parameters of the historical insurance product (such as product code, coverage type, etc.) are extracted. These feature parameters are then used to perform an exact match retrieval in the pre-determined insurance product database of the intent layer node to locate the corresponding historical insurance product entry. Data is then read from the real-time rate field marked in the entry and directly assigned to the pre-determined insurance rate required to calculate the dynamic insurance rate.

[0075] Example 2, based on the same inventive concept as the adaptive insurance premium calculation method based on dynamic Bayesian networks in the previous examples, such as... Figure 2 As shown, this application provides an adaptive insurance premium calculation system based on dynamic Bayesian networks. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0076] The dynamic Bayesian network construction module 10 is used to dynamically collect multi-source heterogeneous information through a multi-source acquisition interface and construct a dynamic Bayesian network in combination with a preset network structure.

[0077] The real-time risk vector acquisition module 20 is used to acquire real-time risk assessment results through the dynamic Bayesian network, and to retrieve the risk vectorization strategy to process the real-time risk assessment results to obtain real-time risk vectors.

[0078] The risk adjustment coefficient acquisition module 30 is used to evaluate and analyze the real-time risk vector in conjunction with the historical risk database to obtain the real-time risk adjustment coefficient.

[0079] The dynamic insurance premium rate acquisition module 40 is used to adjust the predetermined insurance premium rate based on the real-time risk adjustment coefficient to obtain the dynamic insurance premium rate.

[0080] Furthermore, the system is also used to implement the following functions:

[0081] User behavior information is dynamically collected through the user behavior monitoring interface in the multi-source collection interface; disease epidemic information is dynamically collected through the public health database interface in the multi-source collection interface; policy change information is dynamically collected through the government portal website interface in the multi-source collection interface; and the multi-source heterogeneous information is constructed based on the user behavior information, the disease epidemic information, and the policy change information.

[0082] Furthermore, the system is also used to implement the following functions:

[0083] The multi-source heterogeneous information is preprocessed using a standardized processing strategy to obtain target multi-source information, which is then used as input information for the observation layer nodes in the preset network structure. The preset network structure includes hidden layer nodes storing a predetermined potential risk database and intention layer nodes storing a predetermined insurance product database. The dynamic Bayesian network is constructed based on the observation layer nodes, the hidden layer nodes, and the intention layer nodes.

[0084] Furthermore, the system is also used to implement the following functions:

[0085] The standardized processing strategies include time alignment, missing value imputation, and text structuring transformation.

[0086] Furthermore, the system is also used to implement the following functions:

[0087] Extract a first predetermined risk from the predetermined potential risk types; match a first set of risk factors for the first predetermined risk in the predetermined potential risk database; traverse and filter the target multi-source information based on the first set of risk factors to obtain a first set of factor parameters; analyze the first set of factor parameters to obtain a first real-time risk index for the first predetermined risk; determine whether the first real-time risk index reaches a predetermined index limit, wherein the predetermined index limit is stored in the intent layer node; if the first real-time risk index reaches the predetermined index limit, then add the first mapping relationship between the first predetermined risk and the first real-time risk index to the real-time risk assessment result.

[0088] Furthermore, the system is also used to implement the following functions:

[0089] The predetermined potential risk types include at least user behavior risks, disease infection risks, and policy and regulatory risks.

[0090] Furthermore, the system is also used to implement the following functions:

[0091] The risk vectorization strategy contains a risk type order, and sorts the first mapping relationship based on the risk type order to obtain the real-time risk vector.

[0092] Furthermore, the system is also used to implement the following functions:

[0093] The historical risk vector most similar to the real-time risk vector is obtained by traversing the historical risk database; the dynamic time-normalized distance between the historical risk vector and the real-time risk vector is calculated and normalized to obtain the real-time risk confidence level; the real-time risk confidence level is used as the real-time risk adjustment coefficient.

[0094] Furthermore, the system is also used to implement the following functions:

[0095] Match the historical insurance products corresponding to the historical risk vector, and use the real-time insurance premium rate of the historical insurance products matched in the predetermined insurance product database as the predetermined insurance premium rate.

[0096] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0097] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0098] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. An adaptive calculation method for insurance premium rates based on dynamic Bayesian networks, characterized in that, include: Multi-source heterogeneous information is dynamically collected through a multi-source acquisition interface, and a dynamic Bayesian network is constructed by combining it with a preset network structure. The real-time risk assessment results are obtained through the dynamic Bayesian network, and the risk vectorization strategy is invoked to process the real-time risk assessment results to obtain the real-time risk vector. The real-time risk vector is evaluated and analyzed using a historical risk database to obtain a real-time risk adjustment coefficient. The predetermined insurance premium rate is adjusted based on the real-time risk adjustment coefficient to obtain the dynamic insurance premium rate; This also includes: The standardized processing strategy is invoked to perform standardized preprocessing on the multi-source heterogeneous information to obtain target multi-source information, and the target multi-source information is used as the input information of the observation layer nodes in the preset network structure. The preset network structure includes hidden layer nodes storing a predetermined potential risk database and intention layer nodes storing a predetermined insurance product database. The dynamic Bayesian network is constructed based on the observation layer nodes, the hidden layer nodes, and the intent layer nodes. The process of obtaining real-time risk assessment results through the dynamic Bayesian network includes: Extract the first predetermined risk from the predetermined potential risk types; Match the first set of risk factors for the first predetermined risk in the predetermined potential risk database; Based on the first risk factor set, the target multi-source information is traversed and filtered to obtain the first factor parameter set; The first real-time risk index of the first predetermined risk is obtained by analyzing the first factor parameter set; Determine whether the first real-time risk index has reached a predetermined index limit, wherein the predetermined index limit is stored in the intent layer node; If the first real-time risk index reaches the predetermined index limit, then the first mapping relationship between the first predetermined risk and the first real-time risk index is added to the real-time risk assessment result.

2. The adaptive insurance premium calculation method based on dynamic Bayesian networks as described in claim 1, characterized in that, Multi-source heterogeneous information is dynamically collected through a multi-source acquisition interface, including: User behavior information is dynamically collected through the user behavior monitoring interface in the multi-source acquisition interface. Disease epidemic information is dynamically collected through the public health database interface in the multi-source acquisition interface; Policy change information is dynamically collected through the government portal website interface in the multi-source collection interface; The multi-source heterogeneous information is constructed based on the user behavior information, the disease prevalence information, and the policy change information.

3. The adaptive insurance premium calculation method based on dynamic Bayesian networks as described in claim 1, characterized in that, The standardized processing strategies include time alignment, missing value imputation, and text structuring transformation.

4. The adaptive insurance premium calculation method based on dynamic Bayesian networks as described in claim 1, characterized in that, The predetermined potential risk types include at least user behavior risks, disease infection risks, and policy and regulatory risks.

5. The adaptive insurance premium calculation method based on dynamic Bayesian networks as described in claim 4, characterized in that, The risk vectorization strategy contains a risk type order, and sorts the first mapping relationship based on the risk type order to obtain the real-time risk vector.

6. The adaptive insurance premium calculation method based on dynamic Bayesian networks as described in claim 1, characterized in that, By combining historical risk databases with the evaluation and analysis of the real-time risk vector, a real-time risk adjustment coefficient is obtained, including: The historical risk vector that is most similar to the real-time risk vector is obtained by traversing the historical risk database. The dynamic time-warped distance between the historical risk vector and the real-time risk vector is calculated and normalized to obtain the real-time risk confidence. The real-time risk confidence level is used as the real-time risk adjustment coefficient.

7. The adaptive insurance premium calculation method based on dynamic Bayesian networks as described in claim 6, characterized in that, Match the historical insurance products corresponding to the historical risk vector, and use the real-time insurance premium rate of the historical insurance products matched in the predetermined insurance product database as the predetermined insurance premium rate.

8. An adaptive insurance premium calculation system based on dynamic Bayesian networks, characterized in that, The system is used to implement the adaptive insurance premium calculation method based on dynamic Bayesian networks as described in any one of claims 1-7, and the system comprises: The dynamic Bayesian network construction module is used to dynamically collect multi-source heterogeneous information through a multi-source acquisition interface and construct a dynamic Bayesian network in combination with a preset network structure. The real-time risk vector acquisition module is used to acquire real-time risk assessment results through the dynamic Bayesian network, and to retrieve the risk vectorization strategy to process the real-time risk assessment results to obtain real-time risk vectors. The risk adjustment coefficient acquisition module is used to evaluate and analyze the real-time risk vector in conjunction with the historical risk database to obtain the real-time risk adjustment coefficient; The dynamic insurance premium rate acquisition module is used to adjust the predetermined insurance premium rate based on the real-time risk adjustment coefficient to obtain the dynamic insurance premium rate.

Images