Risk data single reinsurance proportion optimization method and device, equipment and medium
By using quantum feature encoding and risk assessment models, the irrationality caused by manual experience and fixed ratio rules in traditional reinsurance ratio methods is resolved, and dynamic adjustment and precise matching of reinsurance ratios are achieved.
Patent Information
- Application Number
- CN202511800750.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-02
- Publication Date
- 2026-03-17
AI Technical Summary
Traditional reinsurance companies rely on manual experience or fixed ratio rules for their ceding ratios, resulting in unreasonable and inefficient ceding ratios that cannot be dynamically adjusted and are difficult to match with policies with high risk fluctuations.
A quantum risk assessment model is constructed by using quantum feature encoding and key risk feature screening to calculate risk scores and dynamically adjust the reinsurance ratio in conjunction with reinsurance risk preferences.
It achieves objectivity and precision in risk assessment, dynamically matches risks with reinsurance ratios, and improves the efficiency and accuracy of reinsurance ratio allocation.
Smart Images
Figure CN121685162A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of intelligent decision-making, and in particular to a risk data single reinsurance ratio optimization method, device, equipment and medium. BACKGROUND
[0002] The risk data single can also be referred to as a policy, and policy reinsurance refers to the operation of transferring part or all of the insurance business of an insurance company (original insurer) to other insurance companies to bear; for example, the original insurance company transfers part or all of the risks of the insurance business it underwrites to other insurance companies according to the established risk sharing ratio, in order to reduce its claim pressure and stabilize operations.
[0003] At present, the ceding ratio method of traditional reinsurance companies has many shortcomings. On the one hand, the ceding ratio method of some reinsurance companies mainly relies on the subjective judgment of underwriting personnel based on industry experience to determine the ceding ratio, but different underwriting personnel may have large differences in risk assessment of the same policy, resulting in unreasonable ceding ratio. In addition, manual processing of a large number of policies is time-consuming and laborious, and prone to errors. On the other hand, the ceding ratio method of some reinsurance companies is to allocate according to fixed ratio rules or fixed threshold method. However, the ratio method and the fixed threshold method are static rules, and cannot dynamically adjust the ceding ratio according to the real-time risk changes of the policy. For policies with large risk fluctuations, the situation of mismatch between reinsurance and risk may easily occur.
[0004] Therefore, in the face of the growing demand for policy reinsurance ratio allocation, the current ceding ratio method of reinsurance companies needs to be improved to solve the problems of low efficiency and insufficient accuracy of ratio allocation of existing methods. SUMMARY
[0005] The present application provides a risk data single reinsurance ratio optimization method, device, equipment and medium, which provides objective and comparable numerical basis for reinsurance ratio allocation by calculating risk scores, and realizes dynamic adjustment of risk data single reinsurance ratio by combining reinsurance risk preference.
[0006] In a first aspect, a risk data single reinsurance ratio optimization method is provided, comprising: Obtaining an original risk data single of a preset risk type, performing quantum feature encoding on the original risk data single to obtain a data quantum state encoding; Performing key risk feature screening on the data quantum state encoding to obtain a key risk feature; Constructing a quantum risk assessment model according to the key risk feature; Calculating a risk score of the original risk data single according to the quantum risk assessment model; Dynamically allocating a reinsurance ratio to the original risk data single according to the risk score and a preset reinsurance risk preference, and forming a reinsurance ratio scheme.
[0007] In a second aspect, a risk data single reinsurance ratio optimization device is provided, comprising: An acquisition module is configured to acquire original risk data singles of a preset risk type, perform quantum feature coding on the original risk data singles, and obtain data quantum state coding. A screening module is configured to perform key risk feature screening on the data quantum state coding, and obtain key risk features. A construction module is configured to construct a quantum risk assessment model according to the key risk features. A calculation module is configured to calculate risk scores of the original risk data singles according to the quantum risk assessment model. A distribution module is configured to dynamically distribute reinsurance ratios of the original risk data singles according to the risk scores and a preset reinsurance risk preference, and form a reinsurance ratio scheme.
[0008] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the above-mentioned risk data single reinsurance ratio optimization method when executing the computer program.
[0009] In a fourth aspect, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program, and the computer program implements the steps of the above-mentioned risk data single reinsurance ratio optimization method when executed by a processor.
[0010] In the above-mentioned risk data single reinsurance ratio optimization method, device, computer device, and storage medium, the original risk data singles are acquired and quantum feature coding is performed, which can convert traditional classical risk data into quantum state form, fully utilize quantum superposition, entanglement, and other characteristics, and provide a technical basis for subsequent processing of high-dimensional and complex correlated risk data; the key risk features are screened from the massive quantum state coded risk features, which can eliminate redundant and irrelevant features, reduce the complexity of subsequent model construction, and improve the efficiency and accuracy of risk assessment; the quantum risk assessment model constructed based on the key risk features can fully utilize the advantages of quantum computing in processing nonlinear and high-dimensional risk correlations, and can more accurately depict the essential laws of risk; the risk scores of the original risk data singles are calculated according to the quantum risk assessment model, the quantitative risk scores replace the traditional fuzzy risk level judgment, provide objective and comparable numerical basis for reinsurance ratio distribution, solve the problem of large differences in risk assessment of the same insurance policy by different underwriters, and standardize and unify risk measurement; based on the real-time calculated risk scores, the reinsurance ratio is dynamically adjusted in combination with the preset reinsurance risk preference, the reinsurance ratio of the insurance policy with large risk fluctuations can be updated in real time with the change of the score, and the accurate matching of reinsurance and risk is ensured. BRIEF DESCRIPTION OF DRAWINGS
[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the description of the embodiments of the present application will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.
[0012] Figure 1 is a schematic diagram of an application environment of a risk data single reinsurance ratio optimization method in an embodiment of the present application; Figure 2 is a flowchart of a risk data single reinsurance ratio optimization method in an embodiment of the present application; Figure 3 is a structural schematic diagram of a risk data single reinsurance ratio optimization device in an embodiment of the present application; Figure 4 is a structural schematic diagram of a computer device in an embodiment of the present application; Figure 5 is another structural schematic diagram of a computer device in an embodiment of the present application. DETAILED DESCRIPTION
[0013] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without any creative effort belong to the scope of protection of the present application.
[0014] The risk data single reinsurance ratio optimization method provided in the embodiments of the present application can be applied in, for example, Figure 1In this application environment, the client communicates with the server via a network. The server can obtain original risk data sheets of preset risk types, encode the original risk data sheets using quantum features to obtain data quantum state codes; screen the data quantum state codes for key risk features to obtain key risk features; construct a quantum risk assessment model based on the key risk features; calculate the risk score of the original risk data sheets based on the quantum risk assessment model; and dynamically allocate reinsurance ratios to the original risk data sheets based on the risk score and preset reinsurance risk preferences to form a reinsurance ratio scheme. The reinsurance ratio scheme is then fed back to the client. This invention provides a risk data sheet reinsurance ratio optimization device. For reinsurance ratio scheme business, by calculating risk scores, it provides objective and comparable numerical basis for reinsurance ratio allocation, and by combining reinsurance risk preferences, it realizes dynamic adjustment of the reinsurance ratio of risk data sheets. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will be described in detail below through specific embodiments.
[0015] Please see Figure 2 As shown, Figure 2 A flowchart illustrating a method for optimizing the reinsurance ratio of risk data sheets provided in an embodiment of the present invention includes the following steps: S1. Obtain the original risk data sheet of the preset risk type, and encode the original risk data sheet with quantum features to obtain the data quantum state code.
[0016] In this embodiment of the invention, the original risk data sheet refers to the original data set that records information related to a certain type of risk. It is a detailed description of a specific risk object and includes various attribute data related to the risk, such as an insurance data sheet (including basic information of the insured, health risk data, and insurance-related information). The quantum feature encoding refers to the process of converting the classical data (such as numerical values, categories, etc.) in the original risk data sheet into a quantum state representation that can be processed by quantum computing.
[0017] Specifically, the original data records (original risk data sheets) of a certain preset risk type are collected, and then these classical data are converted into quantum state forms through quantum computing encoding methods, that is, quantum state encoding; where quantum state refers to the core concept in quantum mechanics that describes the state of a quantum system and is used to characterize all the information of a quantum particle or quantum system.
[0018] In the financial scenario, the default risk type is "credit default risk". The original risk data is a company's credit application data (including revenue, liabilities, industry attributes, credit records, etc.). This data is transformed into a quantum state, and the quantum superposition property is used to simultaneously carry the interaction relationship of the company's multi-dimensional financial indicators, providing a data foundation for the subsequent quantum model to assess default risk and optimize credit decisions.
[0019] In this embodiment of the invention, the step of performing quantum feature encoding on the original risk data sheet to obtain data quantum state encoding includes: Multidimensional risk features are extracted from the original risk data sheet to obtain multidimensional risk features; The multidimensional risk features are standardized to generate a standard risk multidimensional feature vector; A variable quantum feature mapping structure is constructed, and the quantum circuit parameters in the variable quantum feature mapping structure are assigned values according to the standard risk multidimensional feature vector to obtain a quantum feature mapping circuit with complete parameter configuration. The standard risk multidimensional feature vector is converted into a data quantum state code based on the quantum feature mapping circuit and the preset quantum system.
[0020] In this embodiment of the invention, the multidimensional risk feature extraction refers to mining and extracting multiple dimensions of risk-related features from the original risk data sheet. The construction of the variable quantum feature mapping structure refers to designing a bridge structure that combines classical data and quantum computing. This structure includes preset quantum circuit modules (such as rotation gates and entanglement gate sequences) and adjustable parameters to convert standard multidimensional feature vectors into quantum states. Its core is to enable the quantum state to efficiently carry the complex correlations between features through variational optimization. The quantum feature mapping circuit refers to the specific implementation of the variable quantum feature mapping structure, which is composed of a series of quantum gates connected according to specific logic. The circuit contains parameters to be assigned (such as rotation angles). Through parameter configuration, the input feature vector can be mapped to the corresponding quantum state. The conversion refers to the quantum feature mapping circuit with the standard risk multidimensional feature vector input parameters configured. Through the operation of the quantum circuit, the information of the feature vector is written into the quantum system (such as the superposition state or entanglement state of qubits), and finally a quantum state representation that can be directly processed by quantum computing is formed.
[0021] Specifically, when extracting multidimensional risk features from the original risk data sheet, based on the risk assessment objective, feature dimensions that are strongly correlated with the risk are selected, and the selected features are processed in a structured manner (such as retaining the original values of numerical features and converting categorical features into coded values) to form a risk feature set containing multiple dimensions.
[0022] Furthermore, for a risk feature set containing multiple dimensions, the differences in the units of measurement of different dimensions are eliminated. Standardization methods (such as mean-standard deviation standardization, min-max normalization) are used to transform the features of each dimension to a uniform numerical range (such as [-1,1] or [0,1]). The standardized features are arranged in dimensional order to form a vector form (such as [feature1, feature2,..., featuren]), and the standard risk multidimensional feature vector is output.
[0023] Furthermore, when constructing the variable quantum feature mapping structure, the number of qubits in the quantum circuit is determined based on the dimension of the standard feature vector (e.g., n-dimensional) (usually not less than the feature dimension; for example, if n=5, then 5 qubits are required). Subsequently, the basic architecture of the quantum circuit is designed, including a "feature input layer" (which converts feature values into qubit rotation angles through rotation gates), an "entanglement layer" (which establishes correlations between qubits through CNOT gates, etc.), and a "variable parameter layer" (which reserves adjustable rotation angle parameters to optimize the mapping effect). This allows the quantum state to retain the key information of the features to the greatest extent, outputting the variable quantum feature mapping structure.
[0024] Furthermore, when assigning values to the quantum circuit parameters in the variational quantum feature mapping structure according to the standard risk multidimensional feature vector, the values of each dimension of the standard feature vector are allocated to the feature input layer parameters in the circuit according to preset rules (determined through historical experience) (such as using the feature value directly as the initial angle of the rotating gate). Combined with historical data, initial values are assigned to the adjustable parameters of the variational parameter layer (such as the approximate optimal value found through classical optimization algorithms). The integrity of the circuit parameters is checked (ensuring that the parameters of all quantum gates are assigned values), forming a quantum circuit that can be run directly, that is, a quantum feature mapping circuit with complete parameter configuration.
[0025] Furthermore, when the standard risk multidimensional feature vector is converted into a data quantum state encoding according to the quantum feature mapping circuit and the preset quantum system, the quantum feature mapping circuit is loaded into the preset quantum system (existing technology, not described in detail here) (such as a quantum simulator or a real quantum computer), and the quantum system operation circuit is started. Specifically, the operation is carried out by sequentially operating the quantum gates (such as adjusting the state of the qubit by rotating the gate and establishing the correlation by entanglement gate), so that the information of the standard feature vector is encoded into the quantum state of the qubit. After the operation is completed, the state of the qubit of the quantum system is the quantized representation of the corresponding feature vector. The quantum state is recorded (usually in the form of probability amplitude or density matrix), and the quantum state is the data quantum state encoding.
[0026] S2. Screen the key risk features of the data quantum state encoding to obtain key risk features.
[0027] In this embodiment of the invention, the key risk feature screening refers to the process of identifying and extracting the features that have the greatest impact on risk assessment and best reflect the core risk patterns from data that has been converted into quantum state encoding.
[0028] In fintech scenarios, data quantum state encoding may contain a company's multidimensional financial characteristics (such as the quantum representation of revenue growth rate, debt-to-equity ratio, credit history, industry volatility coefficient, etc.). From these encodings, the core characteristics that have the greatest impact on credit default risk (such as "number of overdue payments in the past 6 months", "current ratio", "revenue stability during industry downturns") can be extracted. These core characteristics can directly affect the company's probability of default.
[0029] In this embodiment of the invention, key risk features are screened for the data quantum state encoding to obtain key risk features, including: The hot start parameters are obtained, and the parameters of the preset quantum approximation optimization algorithm are adjusted according to the hot start parameters to form a parameter quantum approximation optimization algorithm framework with configured hot start parameters. The quantum state encoding is evaluated for feature importance based on the aforementioned parameter quantum approximation optimization algorithm framework to obtain a feature importance score; Key risk features are selected from the quantum state encoding based on the feature importance score.
[0030] In this embodiment of the invention, the hot-start parameters refer to the initial parameters obtained from historical data, classical algorithm pre-training results, or optimization experience of similar problems. These parameters are close to the optimal solution range and can help the quantum algorithm skip the inefficient exploration stage of random initialization and quickly converge to a better state. The parameter adjustment refers to the initial assignment or correction of the adjustable parameters (such as the rotation angle in the quantum circuit, the step size of the optimization iteration, etc.) in the preset quantum approximation optimization algorithm, so that the algorithm framework starts running from a better starting point, improving efficiency and accuracy. The feature importance assessment refers to the calculation of the degree of influence of each risk feature in the quantum state encoding on the risk assessment result through the configured parameter quantum approximation optimization algorithm.
[0031] Among them, the quantum approximation optimization algorithm is a hybrid algorithm that combines quantum computing and classical optimization to approximate solutions to combinatorial optimization problems.
[0032] Specifically, optimization results (such as feature weights and optimal iteration parameters) of classical algorithms (such as machine learning models) or convergence parameters of quantum algorithms in similar risk problems are extracted from historical risk assessment data and used as hot-start parameters. When adjusting the parameters of the preset quantum approximation optimization algorithm according to the hot-start parameters, the parameter structure of the preset quantum approximation optimization algorithm is analyzed to clarify the adjustable parameters that need to be initialized in the algorithm (such as the rotation angle range of the quantum gate, the initial step size of the iterative optimization, etc.). Then, the hot-start parameters are mapped to the adjustable parameters of the quantum algorithm (such as converting the feature weights of the classical model into the initial rotation angle of the quantum circuit), and the parameter configuration of the algorithm framework is completed, resulting in the parameter quantum approximation optimization algorithm framework with configured hot-start parameters.
[0033] Furthermore, when evaluating the feature importance of quantum state encoding based on the parametric quantum approximation optimization algorithm framework, the quantum state encoding (containing the quantized representation of multidimensional risk features) is first input into the parametric quantum approximation optimization algorithm framework. Through the evolution and optimization iteration of quantum circuits, the impact of different features on the risk assessment target is simulated. For example, the importance is quantified by "the decrease in model accuracy after removing a feature" or "the strength of the quantum state correlation between the feature and the risk outcome", and the quantitative score (i.e. feature importance score) of each feature is output. The higher the score, the greater the impact of the feature on the risk assessment.
[0034] Furthermore, a screening threshold for key features is set (determined based on historical data experience) (e.g., features ranking in the top 20% of importance scores, or features with scores above a certain threshold). The feature importance scores are compared with the thresholds, and features that meet the score criteria are retained, ultimately resulting in the set of core features most critical for risk assessment.
[0035] S3. Construct a quantum risk assessment model based on the key risk characteristics.
[0036] In this embodiment of the invention, the quantum risk assessment model refers to a risk assessment tool constructed using quantum computing principles (such as quantum superposition and entanglement), which takes key risk characteristics as input and outputs risk quantification results through quantum circuit operations (such as quantum state evolution and measurement).
[0037] Specifically, based on the selected key risk characteristics (i.e. the core characteristics that are most decisive for risk assessment), a risk assessment model using quantum computing technology is designed and constructed. This model processes the key risk characteristics through the logical operations of quantum circuits, and finally achieves a quantitative assessment of the risk.
[0038] In financial scenarios, for example, based on core indicators of "credit default risk" (such as corporate current ratio, number of overdue payments, industry volatility coefficient, etc.), a quantum risk assessment model can be constructed, i.e., a quantum circuit can be designed. Quantum states can be used to carry the correlation information of these characteristics (such as the nonlinear relationship between current ratio and industry volatility). The default probability can be quickly solved through quantum computing, and an assessment result such as "the default risk level of this enterprise is high" can be output, which can assist in credit approval decisions.
[0039] In this embodiment of the invention, constructing a quantum risk assessment model based on the key risk characteristics includes: Classical risk factors are extracted from the key risk features, and feature correlation calculations are performed on the key risk features to generate a quantum entanglement risk matrix; Construct a hybrid quantum classical neural network based on the classical risk factors and the quantum entanglement risk matrix; The quantum weight matrix in the hybrid quantum classical neural network is optimized to obtain a quantum risk assessment model.
[0040] In this embodiment of the invention, the extraction of classical risk factors refers to separating features (such as numerical indicators, categorical codes, etc.) that can be directly processed by classical computing from key risk features. The feature correlation calculation refers to calculating the nonlinear correlation strength between key risk features using quantum computing methods (such as quantum state entanglement measurement) and presenting these correlations in matrix form (matrix elements represent the degree of entanglement between two features), i.e., the quantum entanglement risk matrix. The quantum weight matrix refers to the weight matrix composed of qubit states or quantum gate parameters in a hybrid quantum classical neural network, used to quantify the interaction influence between quantum features. The optimization refers to adjusting the parameters of the quantum weight matrix through classical optimization algorithms (such as gradient descent) or quantum optimization algorithms to minimize the risk assessment error of the hybrid network and finally obtain a stable and accurate model parameter configuration.
[0041] Specifically, when extracting classical risk factors from key risk features, features suitable for classical computation are selected from the key risk features, and their original classical data form is retained as input to the classical network part. When performing feature correlation calculation on key risk features, the quantum state encoding of key risk features is input into the quantum circuit. By measuring indicators such as entanglement entropy and mutual information of quantum states, the quantum entanglement strength between any two features is calculated. Using key risk features as rows and columns, the entanglement strength values are filled into the matrix to form a quantum entanglement risk matrix. The matrix elements reflect the degree of nonlinear correlation of corresponding feature pairs.
[0042] Furthermore, when constructing a hybrid quantum classical neural network based on classical risk factors and the quantum entanglement risk matrix, the extracted classical risk factors are used as input to design classical neural network structures such as fully connected layers and activation functions to handle linear or low-dimensional nonlinear correlations. Based on the quantum entanglement risk matrix, quantum circuit layers (such as quantum layers containing CNOT gates and rotation gates) are designed to take quantum entanglement correlations as input and realize feature interaction through quantum weight matrices. A fusion layer is added at the back end of the network to integrate the output of the classical network with the measurement results (converted into classical data) of the quantum network (such as splicing and weighted summation) to form the final risk assessment output layer, which outputs the hybrid quantum classical neural network.
[0043] Furthermore, when optimizing the quantum weight matrix in the hybrid quantum classical neural network, initial parameters (such as those based on hot-start parameters) are assigned to the quantum weight matrix in the hybrid quantum classical neural network, labeled data (such as known risk outcomes) is input, the risk assessment prediction value of the network is obtained through forward propagation, the error between the prediction value and the true value is calculated, and the parameters of the quantum weight matrix are adjusted using an optimization algorithm (such as classical backpropagation combined with quantum gradient estimation) to minimize the error. The training is repeated until the error converges, and the hybrid network with optimized parameters is obtained, that is, the quantum risk assessment model.
[0044] In this embodiment of the invention, the step of performing feature correlation calculation on the key risk features to obtain the quantum entanglement risk matrix includes: The key risk features are preprocessed using quantum quantization to obtain the processed features; Each of the processed features is converted into a quantum state using preset quantum encoding conditions; By jointly measuring the quantum states of any two of the processed features, the quantum entanglement strength characterizing the quantum correlation strength of the two features is obtained; A quantum entanglement risk matrix is constructed based on the key risk characteristics and the quantum entanglement strength.
[0045] In this embodiment of the invention, the quantum preprocessing refers to the preprocessing of key risk features to adapt them to the requirements of quantum computing; specifically, it includes cleaning feature data, unifying dimensions, and simplifying redundant information. The joint measurement refers to the coordinated measurement operation of the quantum states of two features; after establishing a correlation between the two quantum states through the entanglement gate in the quantum circuit, their quantum states (such as probability distribution and phase relationship) are measured to quantify the degree of mutual influence between the two at the quantum level. The quantum entanglement strength refers to the quantitative index obtained through joint measurement that characterizes the degree of correlation between the two feature quantum states.
[0046] Specifically, when performing quantum preprocessing on key risk features, the first step is to collect the selected key risk features, such as "loan overdue days" and "corporate profit margin" in financial scenarios. Outliers in the features are removed and missing values are filled to ensure data integrity. The values of different features are mapped to the range applicable to quantum encoding. Redundant dimensions with weak correlation to risk in the features are removed, and core information is retained to obtain the processed features.
[0047] Furthermore, each feature in the processed features is converted into a quantum state through preset quantum encoding conditions. The required number of qubits is determined according to the number of processed features (e.g., n features correspond to n qubits). According to the encoding conditions, the value of each processed feature is converted into the quantum state of the corresponding qubit, so that the quantization information of the feature is converted into the probability amplitude or phase of the quantum state, and the quantum state corresponding to each processed feature is obtained.
[0048] Furthermore, when performing a joint measurement on the quantum states of any two features among the processed features, select the quantum states corresponding to any two features from the output quantum states, and entangle the two quantum states by using an entanglement gate (such as a CNOT gate, with the quantum state of feature A as the control bit and the quantum state of feature B as the target bit) to form a joint quantum state. Obtain the entanglement entropy (the lower the entropy value, the stronger the entanglement) or mutual information (the higher the value, the closer the correlation) of the joint quantum state through a quantum measurement operation (such as a projection measurement). Convert the measurement result into a value characterizing the correlation strength between the two (such as a normalized value between 0 and 1), i.e., the quantum entanglement strength.
[0049] Furthermore, when constructing the quantum entanglement risk matrix based on the key risk features and the quantum entanglement strength, the key risk features are used as the rows and columns of the matrix (the rows and columns correspond to the same set of features, and the order is consistent). The quantum entanglement strength between each pair of features is filled into the matrix according to the corresponding position (e.g., the entanglement strength between feature i and feature j is filled into the i-th row and j-th column). Usually, the main diagonal of the matrix (the correlation between features themselves) is assigned a value of 1 (maximum correlation strength) to ensure the integrity of the matrix, thus obtaining the quantum entanglement risk matrix.
[0050] S4. Calculate the risk score of the original risk data sheet according to the quantum risk assessment model.
[0051] In this embodiment of the invention, the calculation refers to the process of inputting the original risk data into the constructed quantum risk assessment model, and then having the model output a quantified risk score through the collaborative processing of internal quantum circuit operations (such as quantum state evolution and entanglement interaction) and classical computational logic (such as feature weighting and result integration).
[0052] Specifically, the initial collection of raw risk data (containing various types of raw information related to risk) is input into the constructed quantum risk assessment model. Through the collaborative operation of quantum and classical computing within the model, a quantitative score (such as risk probability, risk level value, etc.) is finally obtained to measure the risk level corresponding to the raw data, thereby intuitively reflecting the degree of risk of the risk object.
[0053] In this embodiment of the invention, calculating the risk score of the original risk data sheet based on the quantum risk assessment model includes: The quantum entanglement risk matrix is subjected to quantum state evolution processing based on the quantum risk assessment model to generate processed evolutionary quantum state information; The risk quantification value is obtained by linearly weighting the classical risk factor with the evolutionary quantum state information. The risk quantification value is processed non-linearly to obtain the risk score of the original risk data sheet.
[0054] In this embodiment of the invention, the quantum state evolution processing refers to the time evolution of the quantum state corresponding to the quantum entanglement risk matrix in the quantum risk assessment model through a preset quantum circuit (containing quantum gate operations, such as rotation gate and entanglement gate), so that the quantum state changes dynamically according to the correlation law between features, thereby extracting deeper risk correlation information. The linear weighted calculation refers to the linear combination of classical risk factors (numerical key features) and evolved quantum state information (converted into classical values through measurement) according to preset weights (such as "classical factor × weight 1 + quantum state information × weight 2"), integrating the basic risk contribution of the two types of information to obtain a preliminary risk quantification result. The nonlinear processing refers to applying a nonlinear transformation (such as through activation function, exponential function, etc.) to the risk quantification value obtained by linear weighting to capture the nonlinear relationship between risk and features.
[0055] Specifically, when performing quantum state evolution processing on the quantum entanglement risk matrix according to the quantum risk assessment model, the quantum entanglement risk matrix (reflecting the quantum correlation between key risk features) is transformed into the corresponding initial quantum state. First, the quantum circuit module in the quantum risk assessment model is activated, and the initial quantum state is evolved through a preset quantum gate sequence (such as CNOT gate to strengthen entanglement and R_y gate to adjust the quantum state phase). This allows the quantum state to be updated gradually as the circuit runs, deeply mining the hidden risk modes in the feature correlation. After the evolution is completed, the quantum state is measured to obtain classical numerical information (i.e., evolved quantum state information) that can be used for subsequent calculations, retaining the complex correlation features captured by the quantum processing.
[0056] Furthermore, when performing linear weighted calculations on classical risk factors and evolved quantum state information, weights are assigned to classical risk factors and evolved quantum state information respectively based on the degree of influence of evolved quantum state information and extracted classical risk factors on risk (e.g., weight values are determined through model training, with more important information receiving greater weights). The linear weighting formula is then used to calculate the risk quantification value by multiplying the classical risk factors by their corresponding weights and the evolved quantum state information by their corresponding weights, and finally summing the two results.
[0057] Furthermore, when performing nonlinear processing on the risk quantification value, a nonlinear function suitable for the risk assessment scenario is selected (based on historical experience). The risk quantification value is input into the nonlinear function and transformed (such as compressing extreme values and strengthening the nonlinear growth trend of risk). The processed result is standardized (such as mapping to a risk score range of 0-100 points), and finally the risk score corresponding to the original risk data sheet is obtained.
[0058] S5. Based on the risk score and the preset reinsurance risk preference, dynamically allocate the reinsurance ratio to the original risk data to form a reinsurance ratio scheme.
[0059] In this embodiment of the invention, the dynamic allocation of reinsurance ratios refers to the process of flexibly adjusting the reinsurance ratios of different original risk data sheets (corresponding to specific insured objects or businesses) based on the risk score and the preset risk preferences of reinsurance companies. The reinsurance risk preference refers to the quantitative manifestation of the reinsurance institution's (or original insurance company's) acceptance of risk and its management strategy. It usually includes preset risk thresholds, the maximum risk exposure that can be borne, and reinsurance ratio ranges for different risk levels (such as "when the risk score is higher than 80, the reinsurance ratio shall not be lower than 70%" and "when the risk score is lower than 30, the reinsurance ratio shall not exceed 30%), which is the core basis for guiding the allocation of reinsurance ratios.
[0060] In this embodiment of the invention, the step of dynamically allocating a reinsurance ratio to the original risk data based on the risk score and a preset reinsurance risk preference to form a reinsurance ratio scheme includes: The preset reinsurance risk preference is transformed into a set of risk preference parameters; A quantum optimization model for calculating the reinsurance ratio is constructed based on the risk score and the risk preference parameter set. The optimal allocation scheme is solved by the quantum optimization model to obtain a preliminary re-protection ratio allocation scheme; The preliminary reinsurance ratio allocation scheme is reasonably verified to obtain the reinsurance ratio scheme.
[0061] In this embodiment of the invention, the transformation refers to converting the reinsurance company's qualitative or principle-based risk preferences (such as "the reinsurance ratio for high-risk business shall not be less than 60%" and "the maximum amount to be assumed in a single reinsurance is 5 million") into a set of parameters that can be quantified and input into the model. The quantum optimization model is an optimization model built on the principle of quantum computing. It takes risk scores and risk preference parameter sets as inputs and efficiently searches for the optimal reinsurance ratio that satisfies the risk preference constraints through quantum superposition, entanglement and other characteristics. The solution to the optimal solution refers to running the quantum optimization model through the quantum optimization algorithm to find the reinsurance ratio scheme that minimizes reinsurance costs and makes the risk transfer most reasonable, under the premise of satisfying the constraints of the risk preference parameter set (such as not exceeding the maximum risk assumption limit).
[0062] Specifically, when converting the pre-defined reinsurance risk appetite into a risk appetite parameter set, the reinsurance company's risk appetite rules (such as written regulations and business guidelines) are reviewed, and core constraints are clarified (such as "risk score ≥ 80 points, reinsurance ratio ∈ [70%, 90%]", "reinsurance amount per policy ≤ 10 million", "overall reinsurance cost ratio ≤ 20%"). The risk score range is mapped to specific thresholds (such as a high-risk threshold of 80 points and a medium-risk threshold of 50 points), the ratio requirements are converted into range parameters (such as [70%, 90%]), and the amount limit is converted into numerical parameters (such as 10 million). All quantitative parameters are then organized to form a structured risk appetite parameter set.
[0063] Furthermore, using the risk score and risk preference parameter set of the original risk data sheet as input variables, the objective function to be optimized by the quantum optimization model is defined. The constraints in the risk preference parameter set (such as the reinsurance ratio range and amount limit) are transformed into the constraints of the quantum model to ensure that the output scheme conforms to the reinsurance company's risk preference. A quantum circuit framework containing qubits and quantum gates (used to handle parameter correlation) is designed so that the model can search for the optimal reinsurance ratio through quantum computing and output a quantum optimization model based on risk score and risk preference.
[0064] Furthermore, the risk score and risk preference parameter set are input into the constructed quantum optimization model. Using the D-Wave quantum annealer, within the range that meets the constraints, the model searches for the optimal reinsurance ratio that optimizes the objective function (e.g., finding a 75% reinsurance ratio for high-risk businesses, which satisfies the ratio range constraint and minimizes costs). The quantum state is measured to obtain the classical result, and a preliminary reinsurance ratio allocation scheme is output.
[0065] Furthermore, when reasonably verifying the preliminary reinsurance ratio allocation plan, check whether it fully meets all the constraints of the risk preference parameter set (such as whether the ratio is within the range and whether the amount exceeds the limit); if the preliminary plan has conflicts with actual operation (such as the reinsurance ratio being too high, resulting in low willingness of reinsurance companies to take over), make manual fine adjustments; confirm that the adjusted plan not only meets the model optimization objective but is also feasible, and finally form the reinsurance ratio plan.
[0066] As can be seen, in the above scheme, for the reinsurance ratio scheme business, the original risk data sheet of the preset risk type is obtained, and the original risk data sheet is quantum feature encoded to obtain data quantum state encoding; the data quantum state encoding is screened for key risk features to obtain key risk features; a quantum risk assessment model is constructed based on the key risk features; a risk score of the original risk data sheet is calculated based on the quantum risk assessment model; and a reinsurance ratio is dynamically allocated to the original risk data sheet based on the risk score and the preset reinsurance risk preference to form a reinsurance ratio scheme. By calculating the risk score, an objective and comparable numerical basis is provided for the reinsurance ratio allocation, and by combining the reinsurance risk preference, the dynamic adjustment of the reinsurance ratio of the risk data sheet is realized.
[0067] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
[0068] In one embodiment, a risk data statement reinsurance ratio optimization device is provided, which corresponds one-to-one with the risk data statement reinsurance ratio optimization method described in the above embodiments. For example... Figure 3 As shown, this risk data single reinsurance ratio optimization device includes an acquisition encoding module 101, a screening module 102, a construction module 103, a calculation module 104, and an allocation module 105. Detailed descriptions of each functional module are as follows: The encoding module 101 is used to acquire the original risk data sheet of the preset risk type, and to encode the original risk data sheet with quantum features to obtain the data quantum state encoding. The screening module 102 is used to screen key risk features of the data quantum state encoding to obtain key risk features; Module 103 is used to construct a quantum risk assessment model based on the key risk characteristics; Calculation module 104 is used to calculate the risk score of the original risk data sheet according to the quantum risk assessment model; The allocation module 105 is used to dynamically allocate a reinsurance ratio to the original risk data based on the risk score and the preset reinsurance risk preference, thereby forming a reinsurance ratio scheme.
[0069] In one embodiment, the encoding module 101, when performing quantum feature encoding on the original risk data sheet to obtain data quantum state encoding, is used for: Multidimensional risk features are extracted from the original risk data sheet to obtain multidimensional risk features; The multidimensional risk features are standardized to generate a standard risk multidimensional feature vector; A variable quantum feature mapping structure is constructed, and the quantum circuit parameters in the variable quantum feature mapping structure are assigned values according to the standard risk multidimensional feature vector to obtain a quantum feature mapping circuit with complete parameter configuration. The standard risk multidimensional feature vector is converted into a data quantum state code based on the quantum feature mapping circuit and the preset quantum system.
[0070] In one embodiment, when the screening module 102 performs key risk feature screening on the data quantum state encoding to obtain key risk features, it is used to: The hot start parameters are obtained, and the parameters of the preset quantum approximation optimization algorithm are adjusted according to the hot start parameters to form a parameter quantum approximation optimization algorithm framework with configured hot start parameters. The quantum state encoding is evaluated for feature importance based on the aforementioned parameter quantum approximation optimization algorithm framework to obtain a feature importance score; Key risk features are selected from the quantum state encoding based on the feature importance score.
[0071] In one embodiment, when constructing a quantum risk assessment model based on the key risk characteristics, the construction module 103 is used to: Classical risk factors are extracted from the key risk features, and feature correlation calculations are performed on the key risk features to generate a quantum entanglement risk matrix; Construct a hybrid quantum classical neural network based on the classical risk factors and the quantum entanglement risk matrix; The quantum weight matrix in the hybrid quantum classical neural network is optimized to obtain a quantum risk assessment model.
[0072] In one embodiment, the construction module 103, when performing feature correlation calculations on the key risk features to obtain the quantum entanglement risk matrix, is used to: The key risk features are preprocessed using quantum quantization to obtain the processed features; Each of the processed features is converted into a quantum state using preset quantum encoding conditions; By jointly measuring the quantum states of any two of the processed features, the quantum entanglement strength characterizing the quantum correlation strength of the two features is obtained; A quantum entanglement risk matrix is constructed based on the key risk characteristics and the quantum entanglement strength.
[0073] In one embodiment, when calculating the risk score of the original risk data sheet according to the quantum risk assessment model, the calculation module 104 is used to: The quantum entanglement risk matrix is subjected to quantum state evolution processing based on the quantum risk assessment model to generate processed evolutionary quantum state information; The risk quantification value is obtained by linearly weighting the classical risk factor with the evolutionary quantum state information. The risk quantification value is processed non-linearly to obtain the risk score of the original risk data sheet.
[0074] In one embodiment, when the allocation module 105 dynamically allocates reinsurance ratios to the original risk data based on the risk score and a preset reinsurance risk preference to form a reinsurance ratio scheme, it is used to: The preset reinsurance risk preference is transformed into a set of risk preference parameters; A quantum optimization model for calculating the reinsurance ratio is constructed based on the risk score and the risk preference parameter set. The optimal allocation scheme is solved by the quantum optimization model to obtain a preliminary re-protection ratio allocation scheme; The preliminary reinsurance ratio allocation scheme is reasonably verified to obtain the reinsurance ratio scheme.
[0075] This invention provides a reinsurance ratio optimization device for risk data sheets. For reinsurance ratio schemes, it acquires original risk data sheets of a preset risk type, encodes the original risk data sheets using quantum features to obtain data quantum state codes, filters the data quantum state codes for key risk features, constructs a quantum risk assessment model based on the key risk features, calculates a risk score for the original risk data sheets using the quantum risk assessment model, and dynamically allocates reinsurance ratios to the original risk data sheets based on the risk score and a preset reinsurance risk preference, forming a reinsurance ratio scheme. By calculating the risk score, it provides an objective and comparable numerical basis for reinsurance ratio allocation, and by combining the reinsurance risk preference, it achieves dynamic adjustment of the reinsurance ratio for risk data sheets.
[0076] Specific limitations regarding the risk data statement reinsurance ratio optimization device can be found in the limitations of the risk data statement reinsurance ratio optimization method described above, and will not be repeated here. Each module in the aforementioned risk data statement reinsurance ratio optimization device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.
[0077] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When executed by the processor, the computer program implements the functions or steps of a risk data single reinsurance ratio optimization method on the server side.
[0078] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When executed by the processor, the computer program implements the client-side functions or steps of a risk data single reinsurance ratio optimization method.
[0079] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Obtain the original risk data sheet of the preset risk type, and encode the original risk data sheet with quantum features to obtain the data quantum state code; The key risk features are obtained by screening the quantum state encoding of the data; A quantum risk assessment model is constructed based on the aforementioned key risk characteristics; Calculate the risk score of the original risk data sheet based on the quantum risk assessment model; Based on the risk score and the preset reinsurance risk preference, the reinsurance ratio is dynamically allocated to the original risk data to form a reinsurance ratio scheme.
[0080] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Obtain the original risk data sheet of the preset risk type, and encode the original risk data sheet with quantum features to obtain the data quantum state code; The key risk features are obtained by screening the quantum state encoding of the data; A quantum risk assessment model is constructed based on the aforementioned key risk characteristics; Calculate the risk score of the original risk data sheet based on the quantum risk assessment model; Based on the risk score and the preset reinsurance risk preference, the reinsurance ratio is dynamically allocated to the original risk data to form a reinsurance ratio scheme.
[0081] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.
[0082] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0083] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.
[0084] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. If any software tools or components other than those of our company appear in the embodiments, they are merely illustrative examples and do not represent actual use. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A method for optimizing the reinsurance proportion of risk data single, characterized in that, The method comprises the following steps: acquiring original risk data of a preset risk type, performing quantum feature coding on the original risk data to obtain data quantum state coding; performing key risk feature screening on the data quantum state coding to obtain key risk features; constructing a quantum risk assessment model according to the key risk features; calculating a risk score of the original risk data according to the quantum risk assessment model; dynamically allocating a reinsurance ratio to the original risk data according to the risk score and a preset reinsurance risk preference to form a reinsurance ratio scheme.
2. The risk data single reinsurance proportion optimization method of claim 1, wherein, The method comprises the following steps: performing multi-dimensional risk feature extraction on the original risk data to obtain multi-dimensional risk features; performing standardization processing on the multi-dimensional risk features to generate a standard risk multi-dimensional feature vector; constructing a variational quantum feature mapping structure, and assigning values to quantum circuit parameters in the variational quantum feature mapping structure according to the standard risk multi-dimensional feature vector to obtain a quantum feature mapping circuit with completed parameter configuration; converting the standard risk multi-dimensional feature vector into data quantum state coding according to the quantum feature mapping circuit and a preset quantum system.
3. The risk data single reinsurance ratio optimization method of claim 1, wherein, The method comprises the following steps: acquiring a hot start parameter, adjusting a preset quantum approximate optimization algorithm according to the hot start parameter to form a parameter quantum approximate optimization algorithm framework with the hot start parameter configured; evaluating the importance of features according to the parameter quantum approximate optimization algorithm framework to obtain a feature importance score; screening key risk features from the quantum state coding according to the feature importance score.
4. The risk data single reinsurance proportion optimization method of claim 1, wherein, The method comprises the following steps: extracting classical risk factors from the key risk features, and performing feature correlation calculation on the key risk features to generate a quantum entangled risk matrix; constructing a hybrid quantum-classical neural network according to the classical risk factors and the quantum entangled risk matrix; optimizing quantum weight matrices in the hybrid quantum-classical neural network to obtain a quantum risk assessment model.
5. The risk data single reinsurance proportion optimization method of claim 4, wherein, The method comprises the following steps: performing quantumization preprocessing on the key risk features to obtain processed features; converting each feature in the processed features into a quantum state through a preset quantum coding condition; jointly measuring the quantum states of any two features in the processed features to obtain quantum entanglement strength representing the quantum correlation strength of the two features; constructing a quantum entangled risk matrix according to the key risk features and the quantum entanglement strength.
6. The risk data single reinsurance proportion optimization method of claim 4, wherein, The method comprises the following steps: performing quantum state evolution processing on the quantum entangled risk matrix according to the quantum risk assessment model to generate processed evolution quantum state information; performing linear weighting calculation on the classical risk factors and the evolution quantum state information to obtain a risk quantization value; The risk quantification value is nonlinearly processed to obtain a risk score of the original risk data sheet.
7. The risk data single reinsurance proportion optimization method of claim 1, wherein, The original risk data sheet is dynamically assigned a reinsurance proportion according to the risk score and a preset reinsurance risk preference, to form a reinsurance proportion scheme, including: The preset reinsurance risk preference is converted into a risk preference parameter set; A quantum optimization model for calculating the reinsurance proportion is constructed according to the risk score and the risk preference parameter set; An optimal distribution scheme solution of the quantum optimization model is solved to obtain a preliminary reinsurance proportion distribution scheme; The preliminary reinsurance proportion distribution scheme is reasonably verified to obtain the reinsurance proportion scheme.
8. A risk data single reinsurance ratio optimization apparatus, characterized by, It includes: An acquisition module is configured to acquire original risk data sheets of a preset risk type, and perform quantum feature coding on the original risk data sheets to obtain data quantum state coding; A screening module is configured to screen key risk features from the data quantum state coding to obtain the key risk features; A construction module is configured to construct a quantum risk assessment model according to the key risk features; A calculation module is configured to calculate a risk score of the original risk data sheet according to the quantum risk assessment model; A distribution module is configured to dynamically assign a reinsurance proportion to the original risk data sheet according to the risk score and a preset reinsurance risk preference, to form a reinsurance proportion scheme.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the risk data sheet reinsurance proportion optimization method of any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 9. The computer program is executed by the processor to implement the risk data sheet reinsurance proportion optimization method of any one of claims 1 to 7.