Multi-insurance insurance set recommendation control method, system and equipment based on game algorithm

Abstract

The invention relates to the technical field of insurance service intelligent recommendation and combined decision control, in particular to a multi-insurance insurance set recommendation control method, system and equipment based on a game algorithm, and the method comprises the steps: constructing a user portrait which comprises a risk preference coefficient, a guarantee demand vector and a psychological budget upper limit; generating an initial candidate strategy set based on the user portrait and historical data; establishing a double-main-body game model of the insurance company and the user, constructing a nonlinear utility function for the user side, and constructing a revenue function with a profit interval constraint for the insurance company side; solving by adopting a constrained Nash equilibrium iterative algorithm to obtain an equilibrium recommendation strategy set; and finally, carrying out guarantee redundancy check and legality verification on the strategy set, outputting N interpretable recommendation results, and carrying out iterative optimization on model parameters according to user feedback. According to the invention, the dynamic balance between the user guarantee demand and the income of the insurance company in the multi-insurance combination recommendation is realized, and the recommendation accuracy, credibility and insurance conversion rate are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent recommendation and portfolio decision control technology for insurance business, and in particular to a method, system and device for recommendation and control of multi-insurance sets based on game theory algorithms. Background Technology

[0002] In insurance marketing and internet recommendation technology, multi-insurance package recommendations are a core means to improve user coverage and insurance company customer value. Current technologies mainly include rule-based preset package recommendations, collaborative filtering-based single-insurance product recommendations, and price-based or simple weighted scoring recommendations.

[0003] However, existing technologies have at least the following drawbacks: Existing recommendation engines mainly recommend or call preset packages for single insurance types, and do not have the ability to dynamically generate multiple insurance types based on user needs and perform overall strategy optimization control. Existing recommendation strategies only consider the matching degree of user characteristics, without setting profit constraints for insurance companies' portfolio strategies, and cannot control profit levels and operational stability during the recommendation process; Users’ premium sensitivity is often judged by linear deduction or single-point budget, which cannot express the true psychological acceptance threshold and elasticity of different price ranges, resulting in a high user rejection rate. The inability to identify overlapping and redundant coverage of similar risk protections in the combination recommendation process can easily lead to the phenomenon of "over-insurance or duplicate insurance recommendations", which undermines users' trust in making decisions. The game decision-making module does not incorporate recommendation algorithms for strategy initialization optimization, resulting in high complexity of combinatorial space search and poor convergence efficiency of game strategies, making it difficult to meet real-time recommendation requirements. The recommendation results lack an interpretable strategy path backtracking mechanism, which fails to meet the requirements of insurance business compliance review and regulatory strategy credibility verification. Summary of the Invention

[0004] In view of this, the purpose of this invention is to propose a multi-insurance set recommendation control method, system and device based on game theory algorithm, so as to solve the technical problems of non-optimal combination, uncontrollable insurance company revenue, inability to model psychological premium threshold, inability to identify redundancy of protection, slow convergence of game solution and uninterpretable backtracking of recommendation results in the recommendation of multi-insurance sets.

[0005] To achieve the above objectives, this invention provides a multi-insurance set recommendation and control method based on game theory algorithms, comprising the following steps: Step S1: Collect user data and generate a quantified user profile based on the risk preference coding algorithm and the demand vector construction algorithm; Step S2: Based on historical data and generated user profiles, generate an initial candidate insurance combination strategy space through user similarity matching and data mining; Step S3: Using the candidate insurance product combination strategy space as the game strategy set, construct a two-subject game model with users and insurance companies as the two sides of the game. Here, a nonlinear utility function is constructed for the user side to quantify the utility of each combination strategy for the user, and a profit-constrained payoff function is constructed for the insurance company side to quantify the payoff of each combination strategy for the insurance company. Step S4: In the game model, with the insurance company as the leader and the user as the follower, a constrained Nash equilibrium iterative solution algorithm is used to solve the problem. The constraints include the profit constraint and the user's core protection threshold. The solution yields a set of equilibrium recommendation strategies that simultaneously satisfy the goals of both parties. Step S5: Perform redundancy conflict verification and legality verification on the balanced recommendation strategy set, filter and sort, output the recommended combination, and output the strategy path information used to explain the recommendation results. Step S6: Based on the user's feedback behavior on the recommendation results, adaptively update the user profile parameters and game model parameters for subsequent recommendation iterations.

[0006] Preferably, in step S2, the method of generating the initial candidate insurance combination strategy space through user similarity matching and data mining adopts one or more of collaborative filtering, vector recall, or matrix factorization.

[0007] Preferably, the user profile includes a risk preference coefficient, a protection need vector, and a psychological budget ceiling. Preferably, in step S1, the risk preference encoding algorithm uses Gaussian kernel mapping and logistic regression classifier to nonlinearly encode the user risk feature vector to generate a continuously learnable risk preference coefficient, which is used to adjust the weight of the risk item in the user utility function in step S3.

[0008] Preferably, in step S1, the demand vector construction algorithm generates the guarantee demand vector by fusing the user's explicit rating with the implicit rating calculated based on collaborative filtering, and uses adaptive weights related to the sparsity of the guarantee dimension for fusion.

[0009] Preferably, in step S3, the profit-constrained revenue function includes premium income, actuarial costs, and expected claims, and its output is constrained within a preset profit range using a revenue range projection algorithm.

[0010] Preferably, in step S4, the constrained Nash equilibrium iterative solution algorithm performs optimal response calculations including the profit constraint and the guarantee threshold constraint in each iteration, and judges the convergence of the algorithm by jointly using cosine similarity and loss change.

[0011] Preferably, after generating the initial candidate insurance combination strategy space in step S2, the method further includes a strategy pruning step: by calculating the coverage correlation matrix between insurance types and identifying dominating strategies, the candidate combinations are pruned to control the size of the game strategy set.

[0012] Preferably, the adaptive update in step S6 uses a combination of the Bayesian posterior update algorithm and the gradient feedback backpropagation algorithm to jointly update the parameters in the risk preference encoding algorithm and the utility function.

[0013] This invention also provides a multi-insurance set recommendation and control system based on game theory algorithms, comprising: The user profile modeling module is used to collect user data and generate quantified user profiles based on risk preference coding algorithms and demand vector construction algorithms. The initial strategy candidate generation module is used to generate an initial space of candidate insurance combination strategies based on historical data and generated user profiles through user similarity matching and data mining. The game decision control module is used to construct a two-subject game model with the user and the insurance company as the two sides of the game, using the candidate insurance product combination strategy space as the game strategy set, and to perform Stackelberg game equilibrium solution under constraints. The recommendation execution module is used to perform redundancy conflict verification and legality verification on the balanced recommendation strategy set, filter and sort the recommendations, and output the recommended combinations. At the same time, it outputs the strategy path information used to explain the recommendation results. The feedback iterative learning module is used to update the system model parameters based on user feedback behavior.

[0014] The present invention also provides an electronic device, comprising: processor; Memory, used to store computer programs; When the processor executes the computer program, it implements the above-described multi-insurance set recommendation and control method based on game theory algorithm.

[0015] The beneficial effects of this invention are: 1. This invention is the first to propose a game-theoretic decision-making control framework between leaders (insurance companies) and followers (users) in insurance recommendations and apply it to solving multi-insurance combination recommendation strategies. It proposes a learnable, controllable, and nonlinearly modelable user utility function that truly expresses users' risk preferences, premium psychological acceptance thresholds, and changes in marginal utility of protection satisfaction. It also proposes a protection redundancy identification and over-insurance penalty control mechanism to avoid duplicate protection recommendations and improve the trustworthiness of the recommendation scheme. 2. This invention accurately characterizes user psychological preferences by designing a learnable nonlinear user utility function and a redundancy recognition mechanism, avoiding excessive or repeated recommendations and improving the credibility and personalization of the solution. 3. This invention forms an optimized closed loop of profiling, game theory, recommendation, and feedback, achieving rapid strategy convergence and adaptive parameter updates. While meeting users' protection needs, it ensures stable revenue for insurance companies and significantly improves recommendation efficiency and insurance conversion rate. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the multi-insurance set recommendation and control method based on game theory algorithm according to an embodiment of the present invention; Figure 2 This is a block diagram of a multi-insurance set recommendation and control system based on a game theory algorithm, according to an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0019] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] Example 1: like Figure 1 As shown in the embodiments of this specification, a multi-insurance set recommendation and control method based on game theory algorithms is provided, including the following steps: Step S1: Collect user data and generate a quantified user profile based on the risk preference coding algorithm and the demand vector construction algorithm. The user profile includes the risk preference coefficient, protection demand vector, and psychological budget limit. To avoid relying on manual labeling in traditional recommendations, this embodiment employs a multi-feature nonlinear fusion coding algorithm to generate a risk preference coefficient, which serves as a parameter in the game theory model. The risk preference coefficient quantifies a user's subjective tolerance or aversion to risk. The user data input to the algorithm includes a feature vector composed of the user's historical claim frequency, age risk level, occupational risk coefficient, and health risk indicators. The algorithm uses a Gaussian kernel and a logistic regression classifier for nonlinear risk space projection coding.

[0021] Risk preference coefficient RP u The calculation formula is: RP u = σ ( W RP Φ ( x u )+ b RP ).

[0022] in x u The input to the algorithm can be represented as a feature vector consisting of historical claims frequency, age risk level, occupational risk coefficient, and health risk indicators. Φ ( ) represents the Gaussian kernel eigenmap function. ; Let represent the center vector of the Gaussian kernel, and s be the bandwidth parameter of the Gaussian kernel, which controls the smoothness of the mapping. Representing vectors x and The Euclidean distance between them.

[0023] W RP , b RP These are learnable parameters, namely the weight vector and the bias term. These parameters are obtained through supervised learning using historical data, rather than being manually set. σ ( ) represents the Sigmoid function, used to compress the result of a linear combination into a stable numerical range. In this embodiment, the output is adjusted to [0.5, 3] to serve as a stable game response parameter.

[0024] The algorithm outputs the risk preference coefficient. RP u As a risk preference game response parameter of user utility function, it mathematically guarantees the continuity and learnability of user risk characterization, and solves the problems of existing recommendations using discrete level labels, which lead to non-differentiable game response space, non-self-learning, and inaccurate modeling of psychological risk preferences.

[0025] The demand vector construction algorithm is used to calculate the guarantee demand vector, which quantifies the intensity of a user's demand for each specific type of guarantee and serves as a parameter in the game theory model. In this embodiment, the demand vector construction algorithm integrates explicit user feedback and implicit behavior, generating a multi-dimensional guarantee demand vector through adaptive weights. The calculation formula is as follows: ; in This represents the vector of a user's protection needs in the k-th protection dimension. For explicit ratings, data is obtained through user questionnaires, self-input, and other methods. The implicit rating is derived through a user-based collaborative filtering algorithm.

[0026] ; in, The set of users most similar to user u. For the similarity between users u and v, This represents the insurance preferences of similar users v for the k-th type of coverage.

[0027] This represents the adaptive fusion weight, the value of which is determined by the sparsity of the k-th guaranteed dimension data, i.e.: , This indicates the sparsity of the k-th type of guarantee in the user's historical data.

[0028] The psychological budget ceiling refers to the maximum premium a user is subjectively willing to pay for insurance coverage. It can be obtained through questionnaires, user input, analysis based on user historical behavior, and estimation based on similar groups.

[0029] Step S2: Based on historical data and generated user profiles, an initial candidate insurance product combination strategy space is generated through user similarity matching and data mining. Specifically, the method of generating the initial candidate insurance product combination strategy space through user similarity matching and data mining adopts one or more of collaborative filtering, vector recall, or matrix factorization.

[0030] Collaborative filtering calculates the similarity between the current user profile and the historical user profile, and aggregates and deduplicates the insurance combinations actually purchased by similar historical user groups to form a candidate set. Vector recall uses an embedding model to map users and insurance products to the same vector space, and quickly retrieves the insurance product or combination template most relevant to the current user vector through approximate nearest neighbor search. The matrix decomposition is based on the latent features obtained from the user-insurance type interaction matrix decomposition. This predicts the current user's preference for each type of insurance and selects high-preference insurance types to construct a portfolio candidate set.

[0031] The above methods can be used individually or in combination to generate a scalable, personalized, and high-quality initial strategy set. . Step S3: Using the candidate insurance product combination strategy space as the game strategy set, construct a two-subject game model, wherein a nonlinear utility function is constructed for the user side to quantify the utility of each combination strategy for the user, and a profit-constrained payoff function is constructed for the insurance company side to quantify the payoff of each combination strategy for the insurance company. Specifically, the non-linear utility function on the user side is:

[0032] ; in, This indicates the degree of coverage matching that reflects the user's protection needs. This represents the risk-preference return that reflects the user's risk preference coefficient. This represents the non-linear psychological premium cost that reflects the upper limit of the psychological budget. This indicates a redundancy penalty. This indicates a correction of psychological trust. , , , , All are weights.

[0033] The profit-constrained return function on the insurance company's side is: =Premium ; in For insurance companies to combine insurance products Expected returns, Premium Indicates a combination of insurance products Total premium income, Premium = , The pricing premium for insurance type c, For combination Actuarial and operating costs, , This represents the fixed operating costs of insurance type c. Indicate combination The total sum insured provided under the k-th type of protection, Add a cost factor to the risk; For combination The expected compensation amount , This represents the expected payout amount for the k-th type of protection.

[0034] To ensure the stable operation of insurance companies, the "Profit Range Projection (PRP) algorithm" is used to limit profits within a preset range: ,in This represents the final available revenue value after profit constraints, where `clamp` is the projection function. This represents the minimum profit threshold for a single recommendation. This is the maximum profit cap for a single recommendation, designed to prevent excessive pricing that could harm user interests or trigger regulatory risks.

[0035] Therefore, the insurance company's profit objective used for game theory decision-making is... .

[0036] Step S4: In the game model, with the insurance company as the leader and the user as the follower, a constrained Nash equilibrium iterative solution algorithm is used to solve the problem. The constraints include the profit constraint and the user's core protection threshold. The solution yields a set of equilibrium recommendation strategies that simultaneously satisfy the goals of both parties. The game model is solved using the Profit-Constrained Nash Convergence (PCNC) algorithm. This algorithm efficiently solves the Stackelberg game equilibrium between insurance companies and users under the dual constraints of profit and coverage threshold. The algorithm's iterative mechanism is as follows: ; in, For optimal reaction solver, Represents a given constraint policy space. Representing the protection weight, during algorithm iteration, the insurance company's strategy proposals are generated or updated in the strategy space. And call the profit function to calculate This ensures that the insurance company's profit targets are met. As for the strategies proposed by the insurance company... The algorithm calls the user's utility function to calculate the utility value and simulates user behavior. That is, among all possible strategies, the user will choose the strategy that maximizes its utility function as the response. Based on the prediction of the user's response pattern, the algorithm selects the strategy that maximizes the company's final profit from all strategies that meet the profit constraint.

[0037] The algorithm iterates until the following two conditions are simultaneously met, outputting a final convergent set of balanced recommendation strategies. C .

[0038] 1. Policy stability: The direction of the policy vector changes very little between two adjacent iterations. ,in and Let represent the combined strategy vectors output after the t-th and t+1-th iterations, respectively. The cosine similarity function measures how close two vectors are in a given direction. The policy convergence threshold; 2. The rate of change of the objective function is lower than the threshold, such as |ΔLoss| < 0.5%, where ΔLoss represents the change in the overall loss function.

[0039] As one implementation method, before solving the game, this method also uses a Candidate Set Pruning (CSP) algorithm to intelligently prune the exponentially growing candidate combinatorial strategy space in order to control computational complexity.

[0040] The core pruning rules include merging highly correlated branches and eliminating branches based on dominance strategies.

[0041] High correlation merging refers to calculating the correlation between any two types of insurance. and Coverage relevance: ; like If the two policies are deemed to have highly overlapping protection functions, the policyholder will either merge or discard the policy with the higher premium. This represents the Pearson correlation coefficient between insurance types a and b in terms of coverage. These are the feature vectors representing the coverage scope of insurance types a and b, respectively.

[0042] Dominant strategy elimination refers to the strategy of dominating B if, for two combined strategies A and B, A dominates B if A has better or equal coverage across all safeguard dimensions and lower or equal costs. This approach keeps the number of candidate combinations K within a calculable range, significantly reducing the computational burden of game solving.

[0043] Step S5: Perform redundancy conflict verification and legality verification on the equilibrium recommendation strategy set, filter and sort, and output the Top-N recommendation combinations. Simultaneously, output strategy path information to explain the recommendation results. This information is generated by backtracking the key parameters and decision logic in the game-theoretic solution process. Specifically, it includes: explaining how the recommended scheme matches the user's protection demand vector and risk preference coefficient, the relationship between its total premium and the user's psychological budget ceiling, and how it satisfies the insurance company's profit constraints; it also quantitatively displays the contribution of each type of insurance within the combination to the user's utility function, and clarifies how the scheme, as a game equilibrium solution, balances the interests of both the user and the insurance company, thus providing a transparent and reliable decision-making basis.

[0044] Step S6: Based on the user's feedback behavior on the recommendation results, adaptively update the user profile parameters and game model parameters for subsequent recommendation iterations.

[0045] To avoid the problem that traditional recommendation models cannot change recommendation parameters due to rejection of feedback, this invention adopts a fusion of Bayesian Posterior Update (BPU) and Gradient Feedback Policy Update (GFPU) to update model parameters.

[0046] Bayesian posterior update is used to update the prior distribution parameters of user profiles. (e.g., risk preference distribution): Where data represents newly observed user feedback data. Let be the prior probability distribution of the parameters, representing the prior knowledge before the update. Indicates that under given parameters Below, we observe the possibilities of the current feedback data. This represents the updated posterior probability distribution.

[0047] Gradient feedback is used to update the utility function or the learnable weights W in the encoding network. , in The gradient is calculated based on user feedback signals. H This represents the learning rate.

[0048] The convergence condition is that the update is considered complete when the rate of change of the key parameter approaches zero (e.g., less than 0.1%).

[0049] The algorithm's implementation mechanism addresses the problem in existing technologies where recommendation parameters cannot be adaptively updated by reverse training using user acceptance / rejection signals. This enables the model to continuously adapt to user preferences and market changes, improving the long-term accuracy of recommendations and user satisfaction.

[0050] Example 2: This embodiment provides a multi-insurance set recommendation and control system based on game theory algorithms, such as... Figure 2 As shown, it includes: The user profile modeling module is responsible for collecting multi-dimensional user data and constructing a precise quantitative profile that includes user risk preference coefficients, structured protection demand vectors, and psychological budget limits through risk preference coding and demand vector fusion algorithms, providing personalized input for subsequent game recommendation. The initial strategy candidate generation module, based on the user profile and historical insurance behavior database, uses data mining methods such as collaborative filtering, vector recall, or matrix factorization to perform user similarity matching and strategy recall, generating a high-quality, personalized initial candidate insurance combination strategy space. The game decision control module is used to construct a two-subject game model with users and insurance companies as the two sides of the game, using the candidate insurance product combination strategy space as the game strategy set, and to perform Stackelberg game equilibrium solution under constraints. In this model, a nonlinear utility function is constructed for the user side and a profit function with profit constraints is constructed for the insurance company side. The module also outputs an equilibrium recommendation strategy set through a constrained Nash equilibrium solution algorithm. The recommendation execution module is used to intelligently identify redundancy in equilibrium strategies, verify business compliance, and review profit constraints. It sorts the results based on a comprehensive score of user utility and company revenue, and outputs Top-N recommendation results. At the same time, by tracing back the key parameters and logic in the game decision-making process, it generates a strategy path explanation report that includes demand satisfaction, insurance type contribution analysis, budget compliance, and game balance explanation, thereby improving the transparency and credibility of the recommendations. The feedback iterative learning module captures user behavior signals such as acceptance and rejection of recommendation results in real time. Through Bayesian update and gradient backpropagation mechanisms, it dynamically optimizes user profile parameters and game model weights to form a closed-loop learning system, enabling the model to continuously and adaptively evolve.

[0051] This system, through modular design, evolves insurance recommendations from a "one-way matching" to a "two-way game" approach. Each module is connected via standardized data interfaces, forming a dynamic closed loop: the profiling module provides precise input for the game, the game module outputs equilibrium strategies, the recommendation module ensures the reliability and usability of the results, and the feedback module continuously optimizes the front-end model parameters. The system features a specially designed strategy pruning and rapid convergence mechanism, supporting high-concurrency real-time recommendations and seamlessly integrating with existing core insurance business systems or internet platforms. While ensuring stable profit margins for insurance companies, it outputs highly personalized, non-redundant, and traceable insurance portfolio solutions, thereby significantly improving insurance conversion rates and long-term customer satisfaction.

[0052] Example 3: This embodiment provides an electronic device, including: a processor; a memory for storing a computer program; when the processor executes the computer program, it implements the multi-insurance set recommendation control method based on game theory algorithm provided in Embodiment 1 above.

[0053] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0054] In the embodiments provided in this application, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling or direct coupling or communication connection may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0055] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0056] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0057] The implementation of all or part of the processes in the methods of the above embodiments can also be accomplished by a computer program product. When the computer program product is run on a terminal device, the terminal device can implement the steps in the various method embodiments described above.

[0058] The embodiments described above are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application 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; and these 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 this application, and should all be included within the protection scope of this application.

Claims

1. A multi-insurance set recommendation and control method based on game theory algorithm, characterized in that, The method includes the following steps: Step S1: Collect user data and generate a quantified user profile based on the risk preference coding algorithm and the demand vector construction algorithm; Step S2: Based on historical data and generated user profiles, generate an initial candidate insurance combination strategy space through user similarity matching and data mining; Step S3: Using the candidate insurance product combination strategy space as the game strategy set, construct a two-subject game model with users and insurance companies as the two sides of the game. Here, a nonlinear utility function is constructed for the user side to quantify the utility of each combination strategy for the user, and a profit-constrained payoff function is constructed for the insurance company side to quantify the payoff of each combination strategy for the insurance company. Step S4: In the game model, with the insurance company as the leader and the user as the follower, a constrained Nash equilibrium iterative solution algorithm is used to solve the problem. The constraints include the profit constraint and the user's core protection threshold. The solution yields a set of equilibrium recommendation strategies that simultaneously satisfy the goals of both parties. Step S5: Perform redundancy conflict verification and legality verification on the balanced recommendation strategy set, filter and sort, output the recommended combination, and output the strategy path information used to explain the recommendation results. Step S6: Based on the user's feedback behavior on the recommendation results, adaptively update the user profile parameters and game model parameters for subsequent recommendation iterations.

2. The multi-insurance set recommendation and control method based on game theory algorithm according to claim 1, characterized in that, In step S2, the method of generating the initial candidate insurance combination strategy space through user similarity matching and data mining adopts one or more of collaborative filtering, vector recall, or matrix factorization.

3. The multi-insurance set recommendation and control method based on game theory algorithm according to claim 1, characterized in that, The user profile includes risk preference coefficient, protection need vector, and psychological budget limit.

4. The multi-insurance set recommendation and control method based on game theory algorithm according to claim 3, characterized in that, In step S1, the risk preference encoding algorithm uses Gaussian kernel mapping and logistic regression classifier to nonlinearly encode the user risk feature vector, generating a continuously learnable risk preference coefficient, which is used to adjust the weight of the risk item in the user utility function in step S3.

5. The multi-insurance set recommendation and control method based on game theory algorithm according to claim 3, characterized in that, In step S1, the demand vector construction algorithm generates the guarantee demand vector by fusing the user's explicit rating with the implicit rating calculated based on collaborative filtering, and uses adaptive weights related to the sparsity of the guarantee dimension for fusion.

6. The multi-insurance set recommendation and control method based on game theory algorithm according to claim 1, characterized in that, In step S3, the profit-constrained revenue function includes premium income, actuarial costs, and expected claims, and its output is constrained within a preset profit range using a revenue range projection algorithm.

7. The multi-insurance set recommendation and control method based on game theory algorithm according to claim 1, characterized in that, In step S4, the constrained Nash equilibrium iterative solution algorithm performs optimal response calculations in each iteration, including the profit constraint and the guarantee threshold constraint, and judges the convergence of the algorithm by jointly using cosine similarity and loss change.

8. The multi-insurance set recommendation and control method based on game theory algorithm according to claim 1, characterized in that, After generating the initial candidate insurance combination strategy space in step S2, the method further includes a strategy pruning step: by calculating the coverage correlation matrix between insurance types and identifying dominating strategies, the candidate combinations are pruned to control the size of the game strategy set.

9. A multi-insurance set recommendation and control system based on game theory algorithms, characterized in that, include: The user profile modeling module is used to collect user data and generate quantified user profiles based on risk preference coding algorithms and demand vector construction algorithms. The initial strategy candidate generation module is used to generate an initial space of candidate insurance combination strategies based on historical data and generated user profiles through user similarity matching and data mining. The game decision control module is used to construct a two-subject game model with the user and the insurance company as the two sides of the game, using the candidate insurance product combination strategy space as the game strategy set, and to perform Stackelberg game equilibrium solution under constraints. The recommendation execution module is used to perform redundancy conflict verification and legality verification on the balanced recommendation strategy set, filter and sort the recommendations, and output the recommended combinations. At the same time, it outputs the strategy path information used to explain the recommendation results. The feedback iterative learning module is used to update the system model parameters based on user feedback behavior.

10. An electronic device, characterized in that, include: processor; Memory, used to store computer programs; When the processor executes the computer program, it implements the multi-insurance set recommendation control method based on game theory algorithm as described in any one of claims 1 to 8.

Images