Methods, devices, equipment and media for generating risk intervention service plans

Abstract

This invention relates to the field of artificial intelligence technology and discloses a method, apparatus, device, and medium for generating risk intervention service solutions. Belonging to the field of artificial intelligence technology and applied to fintech scenarios, the method includes: fusing at least two of the following features from a target user's behavior characteristics, policy characteristics, and interaction characteristics on an insurance platform to obtain fused user characteristics; assessing the churn risk of the target insured user based on the fused user characteristics to obtain the degree of churn risk; and generating a target risk intervention service solution for the target insured user based on the degree of churn risk. This application, based on at least two of the fused user behavior characteristics, policy characteristics, and interaction characteristics, achieves user churn risk assessment and ultimately generates a targeted risk intervention service solution, improving the matching effect between the risk intervention solution and the user's actual needs, thereby improving the conversion efficiency of the risk intervention service solution.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and is applicable to financial technology scenarios. In particular, it relates to a method, apparatus, equipment, and medium for generating risk intervention service solutions. Background Technology

[0002] Traditional methods for generating risk intervention service plans typically analyze policyholder policy data using churn risk assessment models (such as XGBoost networks) to identify potential churn risk users and implement uniform intervention strategies based on business experience. For example, in auto insurance renewal scenarios, fixed threshold rules can be set based on sales staff experience, such as "more than 2 claims in the previous year" or "no login to the auto insurance app within 30 days of the renewal date," triggering SMS reminders or phone calls, and offering generic gifts to try and retain policyholders. However, this method struggles to capture dynamic changes in user behavior, fails to fully consider the dynamic risk characteristics of different individuals within the policy period, and, due to its reliance on rigid business experience, makes it difficult to develop personalized risk intervention plans for different user churn risk situations. This results in a mismatch between the generated risk intervention plans and the actual needs of users, leading to low conversion rates for risk intervention service plans. Therefore, improving the conversion efficiency of risk intervention service plans has become an urgent problem to be solved. Summary of the Invention

[0003] This invention provides a method, apparatus, computer equipment, and medium for generating risk intervention service plans, in order to solve the technical problem of mismatch between the generated risk intervention plans and the actual needs of users, thereby improving the conversion efficiency of risk intervention service plans.

[0004] Firstly, a method for generating risk intervention service plans is provided, including: Obtain target user data for the target insured users; Feature extraction is performed on the target user data to obtain target user features; wherein, the target user features include at least two of the following: user behavior features on the pre-built insurance platform, user policy features, and user interaction features; At least two of the user behavior features, the user policy features, and the user interaction features are fused to obtain fused user features; Based on the integrated user characteristics, the churn risk assessment of the target insured users is performed to obtain the degree of churn risk of the target users; Based on the risk level of churn of the target users, a target risk intervention service plan is generated for the target insured users.

[0005] Secondly, a risk intervention service plan generation device is provided, including: The user data acquisition module is used to acquire target user data of the target insured user. The feature extraction module is used to extract features from the target user data to obtain target user features; wherein, the target user features include at least two of the following: user behavior features on the pre-built insurance platform, user policy features, and user interaction features; The feature fusion module is used to fuse at least two of the user behavior features, the user policy features, and the user interaction features to obtain fused user features. The churn risk assessment module is used to assess the churn risk of the target insured user based on the integrated user characteristics, and to obtain the degree of churn risk of the target user. The intervention service plan generation module is used to generate a target risk intervention service plan for the target insured user based on the target user's churn risk level.

[0006] Thirdly, 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 implement the steps of the above-described risk intervention service scheme generation method.

[0007] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the above-described risk intervention service scheme generation method.

[0008] The aforementioned risk intervention service plan generation method, apparatus, computer equipment, and storage medium enable the following solutions: Target user data of the target insured user is acquired, and at least two of the target user characteristics (including user behavior characteristics, policy characteristics, and user interaction characteristics) are extracted. These characteristics are then fused to obtain fused user characteristics. Based on these fused user characteristics, the churn risk of the target insured user is assessed to determine the degree of churn risk. Finally, a target risk intervention service plan is generated based on the degree of churn risk. In this invention, for scenarios with dynamic changes in insured user behavior and significant individual differences, at least two of the user behavior characteristics, policy characteristics, and interaction characteristics can be extracted to capture dynamic user characteristics within the policy period. The degree of churn risk is then assessed based on the fused user characteristics to generate a risk intervention service plan with individual differences. This effectively avoids the shortcomings of rigid traditional business experience and uniform intervention strategies, significantly improving the matching effect between the risk intervention plan and the user's actual needs, thereby increasing the conversion efficiency of the risk intervention service plan. Attached Figure Description

[0009] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of an application environment for a risk intervention service plan generation method according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a method for generating a risk intervention service plan in one embodiment of the present invention; Figure 3 yes Figure 2 A flowchart illustrating a specific implementation of step S204; Figure 4 yes Figure 3 A schematic diagram of a specific implementation of step S302; Figure 5 yes Figure 3 A flowchart illustrating a specific implementation of step S303; Figure 6 yes Figure 2 A schematic diagram of a specific implementation method for step S205; Figure 7 This is another flowchart illustrating the risk intervention service plan generation method in one embodiment of the present invention; Figure 8 yes Figure 7 A schematic diagram of a specific implementation method for step S702; Figure 9 This is a schematic diagram of a risk intervention service plan generation device in one embodiment of the present invention; Figure 10 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention; Figure 11 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0011] First, let's analyze some of the terms used in this application: Artificial intelligence (AI) is a new branch of computer science that studies, develops, and applies theories, methods, technologies, and systems to simulate, extend, and expand human intelligence. It aims to understand the essence of intelligence and produce intelligent machines that can react in a way similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. AI can simulate the information processes of human consciousness and thought. Furthermore, AI utilizes digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceiving the environment, acquiring knowledge, and using that knowledge to achieve optimal results.

[0012] This application provides a method, apparatus, device, and medium for generating risk intervention service solutions, aiming to solve the mismatch between the generated risk intervention solutions and the actual needs of users, thereby improving the conversion efficiency of risk intervention service solutions.

[0013] The risk intervention service scheme generation method, apparatus, equipment and medium provided in the embodiments of this application are specifically described through the following embodiments. First, the risk intervention service scheme generation method in the embodiments of this application is described.

[0014] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0015] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0016] The policy claim data filtering method provided in this application relates to the field of artificial intelligence technology. This method can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application implementing the policy claim data filtering method, but is not limited to the above forms.

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] The risk intervention service plan generation method provided in this embodiment of the invention can be applied to, for example... Figure 1 In this application environment, the client communicates with the server via a network. The server can obtain target user data of the target insured user and extract at least two of the target user features, including user behavior features, user policy features, and user interaction features. It then fuses these features to obtain fused user features. Based on the fused user features, it assesses the churn risk of the target insured user to obtain the degree of churn risk. Finally, it generates a target risk intervention service plan for the target insured user based on the degree of churn risk. In this invention, for scenarios with dynamic changes in insured user behavior and significant individual differences, at least two of the user behavior features, user policy features, and user interaction features can be extracted to capture dynamic user characteristics within the policy period. The degree of churn risk is then assessed based on the fused user features to generate a risk intervention service plan with individual differences. This effectively avoids the shortcomings of rigid business experience and uniform intervention strategies in traditional methods, significantly improving the matching effect between the risk intervention plan and the user's actual needs, thereby improving the conversion efficiency of the risk intervention service plan. 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.

[0019] like Figure 2 As shown, Figure 2 A flowchart illustrating the risk intervention service solution generation method provided in this embodiment of the invention may include, but is not limited to, steps S201 to S205: Step S201: Obtain target user data for the target insured user.

[0020] Step S202: Extract features from the target user data to obtain target user features; wherein, the target user features include at least two of the following: user behavior features on the pre-built insurance platform, user policy features, and user interaction features.

[0021] Step S203: At least two of the user behavior features, user policy features, and user interaction features are fused to obtain fused user features.

[0022] Step S204: Assess the churn risk of the target insured users based on the integrated user characteristics to obtain the degree of churn risk of the target users.

[0023] Step S205: Generate a target risk intervention service plan for the target insured user based on the risk level of target user churn.

[0024] Steps S201 to S205 of this embodiment involve acquiring target user data of the target insured user, and extracting at least two of the target user features from the target user data, including user behavior features, user policy features, and user interaction features; fusing at least two of the user behavior features, user policy features, and user interaction features to obtain fused user features; assessing the churn risk of the target insured user based on the fused user features to obtain the degree of churn risk of the target user; and generating a target risk intervention service plan for the target insured user based on the degree of churn risk of the target user. In this invention, for scenarios where the behavior of insured users changes dynamically and there are significant individual differences among users, at least two of the user behavior features, user policy features, and user interaction features can be extracted to capture the dynamic features of users within the policy period, and the degree of churn risk of users can be assessed based on the fused user features to generate a risk intervention service plan with individual differences. This can effectively avoid the defects of rigid business experience and uniform intervention strategies in traditional business, significantly improve the matching effect between risk intervention plans and users' real needs, and thus improve the conversion efficiency of risk intervention service plans.

[0025] In step S201 of some embodiments, specifically, target user data refers to the original data set related to the target insured user, including multi-dimensional information such as user behavior data, insurance policy data and interaction data on the pre-built insurance platform.

[0026] Taking auto insurance renewal as an example, the target policyholders can be existing auto insurance customers whose policies are about to expire. The target user data can include the customer's login logs on the auto insurance APP, the current policy's insurance period, premium information and vehicle accident records, call records of the user consulting customer service about the claims process, complaint types, and marketing response rate, etc.

[0027] Specifically, raw data of target insured users can be collected from the database and business systems of the pre-built insurance platform, including user identification information, behavior log data, basic policy data, and interaction record data.

[0028] In this embodiment, by acquiring the target user data of the target insured user, it is possible to collect multi-source data of the target insured user, which provides a complete data foundation for subsequent user feature extraction and avoids the feature loss problem caused by a single data source in traditional methods.

[0029] In step S202 of some embodiments, specifically, the target user feature refers to a set of data vector representations extracted from the target user data. The target user feature includes at least two of the following: user behavior features, user policy features, and user interaction features on the pre-built insurance platform. Among them, the user behavior feature refers to the vector representation of the target insured user's operating habits and active status on the pre-built insurance platform; the user policy feature refers to the vector representation of the target insured user's insurance contract holding status and historical records; and the user interaction feature refers to the vector representation of the interaction between the target insured user and the insurance platform service touchpoints.

[0030] Taking auto insurance renewal as an example, user behavior characteristics may include the number of times the insured user logs into the auto insurance APP in the past 30 days, the length of time spent on the renewal quote page, and the frequency of clicking on the auto insurance claims function; user policy characteristics may include the remaining days of the current policy's insurance period, the number of accidents in the previous year, and the number of historical renewals; user interaction characteristics may include the number of times the user consults the customer service hotline this month, the click status of renewal discount marketing text messages, and the duration of online customer service conversations.

[0031] Specifically, for user policy features, word embedding can be performed on user policy data to extract dialogue vector features; for user behavior and interaction data, behavior and interaction data can be embedded to obtain behavior vector sequences and interaction vector sequences, and then RNN (Recurrent Neural Network) can be used to perform behavior analysis on the behavior vector sequences and interaction vector sequences step by step to obtain time-series-based user behavior features and user interaction features.

[0032] In this embodiment, by extracting at least two of the following from the target user data: user behavior characteristics on the pre-built insurance platform, user policy characteristics, and user interaction characteristics, the scattered raw data can be transformed into target user characteristics with clear business meaning and technical interpretability. It also effectively captures the dynamic changes and individual differences in the behavior of insured users during the policy period, laying a feature-level foundation for the subsequent generation of user characteristics and the accuracy of churn risk assessment, and improving the ability of the entire risk intervention service solution generation method to identify complex user characteristics.

[0033] In step S203 of some embodiments, specifically, fusing user features refers to the comprehensive vector representation formed by integrating at least two of the user behavior features, user policy features, and user interaction features.

[0034] Specifically, the extracted user behavior features, user policy features, and user interaction features can be concatenated to generate fused user features.

[0035] Taking car insurance policyholders as an example, user behavior characteristics such as "login to the car insurance APP 5 times" and "browsing car insurance products for 12 minutes", user policy characteristics such as "1 accident" and "15 days remaining in the policy", and user interaction characteristics such as "2 customer service inquiries" and "no click on push SMS", can be vectorized to form a fused user feature of six-dimensional feature vector. Furthermore, the interaction feature value of "login frequency and remaining days of the policy" can be generated through feature cross-referencing to characterize the platform activity of the target policyholder when the policy is about to expire.

[0036] In this embodiment, by fusing at least two of the user behavior features, user policy features, and user interaction features, a fused user feature is obtained. This can overcome the information limitations of a single user feature type. By capturing the complex patterns of user behavior through the cross-combination of multi-source features, the accuracy of multi-source feature fusion representation in assessing the risk of subsequent user churn is significantly improved, thereby helping to provide risk intervention service solutions with higher conversion rates for subsequent users.

[0037] like Figure 3 As shown, in some embodiments, step S204 includes, but is not limited to, steps S301 to S303: Step S301: Based on the integrated user characteristics, conduct a preliminary churn assessment of the target insured users to obtain the initial user churn risk level.

[0038] Step S302: If the initial user churn risk level meets the predetermined churn risk conditions, then perform risk analysis on the integrated user characteristics based on the initial user churn risk level to obtain the target churn risk characteristics.

[0039] Step S303: Assess the churn risk of the target insured users based on the initial churn risk level and the target churn risk characteristics to obtain the target user churn risk level.

[0040] In step S301 of some embodiments, specifically, the initial user churn risk level refers to a quantitative indicator of the initial predicted churn probability of the target insured user, which is usually presented in the form of a probability value.

[0041] Specifically, the initial churn assessment of target insured users can be achieved based on a pre-trained churn prediction model (such as XGBoost (eXtreme Gradient Boosting) model). This model further includes a split decision layer. Using the initial split feature index, initial split threshold, and initial leaf weights on this split decision layer, a multi-round iterative decision tree is constructed on the fused user features. At each node of each decision tree, the optimal split feature and split threshold are searched using a greedy algorithm to divide the fused user features into left or right child nodes until a leaf node is reached. Each leaf node stores the initial leaf weights for predicting the renewal probability. The ensemble prediction layer in this model performs a weighted summation of each initial leaf weight and its corresponding leaf node. The sigmoid function is then used to activate the weighted summation result, yielding the initial renewal probability of the target insured user, i.e., the initial user renewal risk level of the target insured user.

[0042] Specifically, a preliminary churn assessment of target insured users is conducted based on the integrated user characteristics to obtain the initial user churn risk level. This includes: predicting the renewal of target insured users based on a pre-trained churn prediction model (such as XGBoost (eXtreme Gradient Boosting) model) and integrated user characteristics to obtain the user renewal probability; and calculating the churn probability of target insured users based on the user renewal probability to obtain the initial user churn risk level.

[0043] Taking the car insurance renewal scenario as an example, the initial splitting feature index of the churn prediction model can be used to select "remaining days of the policy" as the root node splitting feature in the first iteration, and the cross feature of "historical loss ratio and marketing response label" as the splitting feature in the second iteration. After 100 iterations, 100 decision trees are constructed. The leaf nodes of each tree store the initial leaf weights (e.g., a leaf node weight of -0.85 indicates that the user represented by this path has a low renewal tendency). In the ensemble output layer, the leaf weights of the 100 trees are weighted and summed using an initial learning rate of 0.05 to obtain an additive output value of -1.25. After Sigmoid compression mapping, the initial renewal probability vector [0.22, 0.78] is obtained, where 0.22 represents the renewal probability and 0.78 represents the non-renewal probability, i.e., the initial user churn risk level.

[0044] like Figure 4 As shown, in some embodiments, step S302 includes, but is not limited to, steps S401 to S402: Step S401: Calculate the risk contribution of the merged user characteristics based on the initial user churn risk level to obtain the characteristic risk contribution degree; wherein, the characteristic risk contribution degree is used to assess the degree to which the merged user characteristics affect the churn risk of the target insured users.

[0045] Step S402: Based on the feature risk contribution degree and predetermined contribution conditions, the fused user features are filtered to obtain the target churn risk features.

[0046] In step S401 of some embodiments, specifically, the feature risk contribution degree refers to a numerical indicator that characterizes the degree and direction of the influence of a single feature in the fused user features on the risk of churn of the target insured user. A positive value indicates that the feature promotes the increase of risk, and a negative value indicates that the feature inhibits the risk.

[0047] In the property insurance scenario, considering the long policy cycle (e.g., 1 year) and significant lapse lag (e.g., signals only appear 3 months before renewal), the characteristic risk contribution rate needs to be able to identify key risk signals that suddenly increase at the end of the policy cycle. Specifically, taking auto insurance renewal business as an example, the characteristic risk contribution rate can quantify the degree to which the characteristic of "15 days remaining in the policy" contributes to the lapse risk as +0.15.

[0048] Specifically, the contribution of characteristic risk can be calculated using the following formula:

[0049] in, S represents the feature risk contribution of feature j in the fused user features, and S represents the feature subset of the fused user features. Let represent the expectation operator, specifically the mathematical expectation of all possible values ​​of the feature subset S, and calculate the average performance of the churn prediction model output under all possible missing feature j scenarios. F represents the fused user features. This represents the set of features remaining after removing feature j from the fused user features. This indicates the level of user churn risk predicted by the model based on feature information, provided that the feature set contains all features in S and feature j. This indicates the degree of user churn risk predicted by the model based on limited feature information when the feature set only contains features in S and does not contain feature j.

[0050] In step S402 of some embodiments, specifically, the predetermined contribution condition refers to the preset screening criteria used to determine whether the fused user characteristics constitute key risk drivers, which is usually determined by the feature risk contribution threshold.

[0051] For example, features whose feature risk contribution is greater than a predetermined feature risk contribution threshold (such as 0.05) can be included in the screening scope.

[0052] Specifically, target churn risk characteristics refer to the set of specific factors that lead to the risk of churn among target insured users, that is, the combination of characteristics that has a significant positive contribution to the degree of initial user churn risk.

[0053] Taking the car insurance renewal scenario as an example, features with a contribution value greater than 0.05 can be selected from the integrated user characteristics to obtain the target churn risk features such as "15 days remaining on the policy" (+0.15), "no click on the renewal discount SMS" (+0.12), and "historical loss ratio" (+0.08).

[0054] Through steps S401 to S402, a complete process from risk quantification to risk attribution and then to risk screening is realized. This effectively overcomes the shortcomings of traditional methods that only output risk results without explaining the causes, or that simple feature importance ranking cannot distinguish between positive and negative contributions. This makes the risk assessment results both predictive and interpretable, providing data-driven decision support for retaining target insured users and improving the conversion efficiency of subsequent risk intervention service plan generation.

[0055] like Figure 5 As shown, in some embodiments, step S303 includes, but is not limited to, steps S501 to S504: Step S501: Filter out the target feature risk contribution of the target churn risk feature from the feature risk contribution.

[0056] Step S502: Calculate the churn risk score for the target insured user based on the risk contribution of the target feature and the initial user churn risk level.

[0057] Step S503: Based on the target churn risk characteristics, predict the value of the target insured users to obtain user value data; wherein, the user value data is used to characterize the expected revenue provided by the target insured users to the insurance platform.

[0058] Step S504: Assess the value risk level of the target insured users based on the user churn risk score and user value data to obtain the degree of target user churn risk.

[0059] In step S501 of some embodiments, specifically, the target feature risk contribution degree refers to the specific contribution degree value corresponding to the target churn risk feature selected from the feature risk contribution degrees.

[0060] Taking the car insurance renewal scenario as an example, we can select from the set of characteristic risk contribution values ​​of the insured user that the target churn risk characteristic "15 days remaining in the policy" corresponds to a target characteristic risk contribution value of +0.15, and the target characteristic risk contribution value "not clicking on the renewal discount SMS" corresponds to a target characteristic risk contribution value of +0.12. These two values ​​constitute the target characteristic risk contribution value of the user.

[0061] In step S502 of some embodiments, specifically, the user churn risk score refers to a numerical indicator that characterizes the overall churn risk level of the target insured user and is used to reflect the specific degree of user churn risk.

[0062] Specifically, the user churn risk score can be calculated using the following formula:

[0063] in, This represents the churn risk score for user i. This indicates the initial user churn risk level for user i. This represents the contribution of feature j in the fused user features to the feature risk of user i; Weighted parameters indicating the degree of churn risk; The weight parameter represents the feature contribution of feature j; This represents the feature weight of feature j.

[0064] In step S503 of some embodiments, specifically, user value data refers to a quantitative indicator that characterizes the expected return provided by the target insured user to the insurance platform, usually presented in the form of predicted premiums and probability of claims or customer lifetime value.

[0065] Specifically, the target churn risk characteristics can be input into a pre-trained customer value assessment model (such as a gradient boosting tree) to predict premiums and claim probabilities, so as to output user value data containing premium and claim prediction probabilities.

[0066] Specifically, the method for predicting the value of target insured users based on their churn risk characteristics to obtain user value data is the same as the method for conducting a preliminary churn assessment of target insured users based on integrated user characteristics, and will not be elaborated here.

[0067] Taking the car insurance renewal scenario as an example, the policy information (compulsory traffic accident liability insurance + commercial insurance combination, last year's premium of 6,200 yuan, no-claims bonus coefficient of 0.85) associated with the target churn risk characteristic "15 days remaining in the policy" and the historical accident record (1 accident in the previous year) can be input into the pre-trained value assessment model to predict the user's next year's premium income of 5,800 yuan and the probability of an accident of 0.6.

[0068] In step S504 of some embodiments, specifically, the target user churn risk level refers to the complete risk assessment result that integrates the user churn risk score and user value data, including the target user process risk level including the user churn risk level and the user value level. This risk level is used to guide the generation of risk intervention service plans. The user churn risk level includes high churn risk level, medium churn risk level and low churn risk level; the user value level includes high value level, medium value level and low value level.

[0069] For example, a user churn risk score greater than 0.70 can be classified as high churn risk, 0.40 to 0.70 as medium churn risk, and less than 0.40 as low churn risk. Users with predicted premium income above 5,000 yuan and a claim probability below 0.3 can be classified as high-value; users with predicted premium income between 3,000 and 5,000 yuan or a claim probability between 0.3 and 0.6 can be classified as medium-value; and users with predicted premium income below 3,000 yuan and a claim probability above 0.6 can be classified as low-value.

[0070] Through steps S501 to S504, the core information of when the risk is high, why the risk is high, and what the value is can be accurately identified during the critical window before policy renewal. This effectively overcomes the shortcomings of traditional single risk scoring, which cannot distinguish customer value and leads to ineffective retention of high-cost customers. It helps to generate subsequent risk intervention service plans that are based on accurate risk attribution and take into account the value of insurance business, thus significantly improving the conversion efficiency of risk intervention service plan generation.

[0071] Steps S301 to S303 effectively overcome the shortcomings of traditional methods in that risk prediction and risk attribution are disconnected, providing a comprehensive and interpretable risk assessment basis for the accurate generation of subsequent risk intervention service plans, and significantly improving the matching degree between risk intervention service plans and users' actual risk situation.

[0072] like Figure 6 As shown, in some embodiments, the risk level of the target user process includes the user churn risk level and the user value level. Step S205 includes, but is not limited to, steps S601 to S603: Step S601: Extract features from user churn risk level to obtain user churn risk level features, and extract features from user value level to obtain user value level features.

[0073] Step S602: Based on the characteristics of user churn risk level, user value level, and target churn risk, an intervention strategy is generated to obtain the risk level intervention strategy.

[0074] Step S603: Generate an intervention service plan based on the risk level intervention strategy indication pre-trained risk intervention generation model to obtain the target risk intervention service plan.

[0075] In step S601 of some embodiments, specifically, the user churn risk level feature refers to a vector representation extracted from the user churn risk level to characterize the level of churn risk.

[0076] Specifically, user value level features refer to vector representations extracted from user value levels that characterize the level of customer value.

[0077] Specifically, word embeddings can be performed on user churn risk level and user value level respectively using a large language model to obtain user churn risk level features corresponding to user churn risk level and user value level features corresponding to user value level.

[0078] In step S602 of some embodiments, specifically, the risk level intervention strategy refers to a strategy instruction that represents the intervention method and the intensity of resource investment, generated by joint decision-making based on user churn risk level characteristics, user value level characteristics and target churn risk characteristics.

[0079] Specifically, the risk intervention instruction information includes the timing of intervention services, the intervention interaction method, and the intervention service level strategy. Based on user churn risk level characteristics, user value level characteristics, and target churn risk characteristics, the risk intervention instruction information can be constructed, which may include: detecting the intervention timing based on target churn risk characteristics to obtain the intervention service timing; identifying interaction preferences based on user churn risk characteristics to obtain the intervention interaction method; and generating a service strategy based on user churn risk level characteristics, user value level characteristics, target churn risk characteristics, and intervention service timing to obtain the intervention service level strategy.

[0080] Taking the car insurance renewal scenario as an example, a risk level intervention strategy can be constructed based on user churn risk level characteristics [1,0,0] (high churn risk level), user value level characteristics [0,1,0] (medium value level), and target churn risk characteristics such as "policy nearing expiration and insensitive to marketing". The strategy is "intervention priority is high priority, intervention interaction channel is dedicated telephone follow-up, discount is an additional 3% discount on car insurance, and intervention timing is 7 days before the policy expires".

[0081] In step S603 of some embodiments, specifically, the pre-trained risk intervention generation model refers to a large language model trained based on historical intervention case data, used to refine risk level intervention strategies into executable service plans.

[0082] Specifically, the target risk intervention service plan refers to a complete set of executable service plans tailored to the specific risk status and value level of the target insured user, including immediate contact service level plans and enhanced intervention service level plans. Among them, the immediate contact service level plan refers to the basic intervention service plan with higher priority and executed at an earlier time in the target risk intervention service plan, which usually includes standardized script templates and basic preferential plans. The enhanced intervention service level plan refers to the upgraded intervention service plan with higher priority and executed at a later time in the target risk intervention service plan, which usually includes personalized scripts, dedicated customer service support, and enhanced preferential plans.

[0083] For example, if the target insured user is a high-value, high-churn auto insurance customer, the instant outreach service level plan can push temporary upgrades to the user's insurance coverage (such as a free 30-day increase in coverage). It can also provide the user with dedicated customer service outbound calls and attribute churn based on user feedback behavior. If the churn is due to "slow insurance claims," ​​an enhanced intervention service level plan can be generated to trigger the claims green channel privilege. If the churn is due to "high premiums," the enhanced intervention service level plan can push personalized discount plans (NCD coefficient adjustment and multi-insurance package discount plans). For users with auto insurance, it can also provide services such as annual free vehicle safety inspections and chauffeur services.

[0084] For example, if the target policyholders are mid-value, high-churn car insurance customers, the instant outreach service level plan can be a combination of intelligent outbound calls and SMS interaction. The specific push content can be renewal reminders and limited-time coupons (such as 200 off for purchases over 2000). Based on user feedback behavior (if there is no response within 3 days), the enhanced intervention service level plan can trigger customer service to follow up again and recommend supplementary insurance products based on user profile (such as adding driver and passenger insurance to increase the overall policy value). If the user feedback behavior indicates successful renewal, the user level is upgraded to enjoy convenient services such as free vehicle inspection the following year.

[0085] For example, if the target insured user is a low-value, high-churn car insurance user, the instant outreach service level plan can be via SMS interaction. The plan includes basic renewal discounts, and if the user has specific risk points in history (such as multiple small claims), further push safe driving courses or disaster prevention and loss mitigation tips to enhance the service. If the user's feedback indicates that they will not renew the policy, the user will be added to the dormant customer pool to reduce the frequency of contact, so as to ensure the service efficiency of the risk intervention service plan.

[0086] For example, if the target insured users are high-value, low-churn home insurance users, the instant outreach service level plan can include strategies such as pushing customer satisfaction surveys, collecting opinions, and giving away phone credit vouchers. Based on the survey feedback data, a policy check report generated for the customer can be pushed to show the coverage gaps and guide users to supplement their insurance.

[0087] For example, if the target insured users are low-value, low-churn users, the instant outreach service level plan can push standardized renewal reminders, while the enhanced intervention service level plan can trigger cross-selling (such as recommending home insurance to car insurance customers).

[0088] Specifically, the risk level intervention strategy can be input as a control command into a pre-trained risk intervention generation model. The encoder in the model encodes the risk level intervention strategy into a latent state vector, and the decoder in the model generates an intervention content sequence based on the latent state vector through autoregression. Through a conditional control mechanism, an immediate access service level plan and an enhanced intervention service level plan are generated respectively.

[0089] Through steps S601 to S603, personalized intervention plans that match the customer's risk level, value level, and specific risk factors can be quickly generated during the critical window before policy renewal, significantly improving the conversion efficiency of risk intervention service plans.

[0090] like Figure 7 As shown, in some embodiments, the target risk intervention service plan includes an immediate reach service level plan and an enhanced intervention service level plan. The risk intervention service plan generation method also includes, but is not limited to, steps S701 to S702: Step S701: Obtain the first time point before the policy expiration date for the target insured user, and push the instant outreach service level plan to the target insured user based on the first time point and the intervention interaction method.

[0091] Step S702: Obtain feedback behavior data of the target insured user after the push. If the feedback behavior data indicates that the user does not have the intention to renew the policy within the preset response period, obtain the second time point of the target insured user before the policy expiration date, and update the intervention interaction method based on the feedback behavior data to obtain the updated intervention interaction method. Based on the updated intervention interaction method and the second time point, push the enhanced intervention service level plan to the target insured user; wherein, the first time point is earlier than the second time point.

[0092] In step S701 of some embodiments, specifically, the instant outreach service level plan refers to the basic intervention service plan with higher priority and executed at an earlier time point in the target risk intervention service plan, which usually includes standardized script templates and basic preferential plans.

[0093] Specifically, the first time point refers to the intervention trigger point set in advance before the policy expiration date for implementing the instant outreach service level plan.

[0094] Specifically, the intervention interaction method refers to the specific channels and forms used to push intervention service plans to the target insured users, including SMS, telephone, APP push, WeChat messages, etc.

[0095] Specifically, the policy expiration date refers to the date on which the insurance contract currently held by the target insured user terminates and becomes invalid.

[0096] Taking car insurance renewal as an example, for high-value and high-risk target policyholders, the policy expiration date can be obtained as May 22, 2024. Based on the high churn risk level, the first time point is set to 30 days before the expiration date, i.e., April 22, 2024. At 10:00 on April 22, 2024, an instant outreach service plan can be pushed to the user via SMS. The specific SMS content is: "Dear Mr. Zhang, your car insurance policy will expire on May 22. The premium for this year is 5,800 yuan, and the no-claims bonus coefficient is 0.85. Click the link to renew your policy with one click and enjoy convenient services. Reply T to unsubscribe." The SMS sending status and delivery receipt are recorded, confirming that the intervention interaction method is SMS and the push status is successful delivery.

[0097] In step S702 of some embodiments, specifically, feedback behavior data refers to the behavioral response data generated by the target insured user after receiving the instant outreach service level plan, which may include click behavior, browsing behavior, inquiry behavior, etc.

[0098] Specifically, the preset response period refers to the time window used to determine whether a customer intends to renew their policy, starting from the moment the service level plan is pushed out.

[0099] Specifically, renewal intention behavior refers to specific actions that indicate the target insured user's willingness to renew their policy within a preset response period, including clicking on a renewal link, browsing a price quote page, or calling customer service.

[0100] Specifically, the second time point refers to the intervention trigger point set in advance before the policy expiration date for implementing the enhanced intervention service level plan, which is later than the first time point.

[0101] Specifically, after pushing out the instant outreach service level plan, feedback behavior data collection is initiated. This involves aggregating user behavior data from multiple sources, including event logs, channel receipts, and records from the insurance business app. At the end of the preset response period, the feedback behavior data is analyzed to determine if the policyholder has any intention to renew. If they click on a renewal link, spend more than a threshold on the quote page, or actively contact customer service, they are considered to have renewal intentions, and the process terminates or proceeds to sales follow-up. If none of these behaviors are present, they are considered not to have renewal intentions, triggering an enhanced intervention process. For the target policyholder triggering enhanced intervention, the policy expiration date can be used to calculate a second time point. Based on the feedback behavior data analysis results, the intervention interaction method is updated, typically upgrading low-response channels to high-response channels (e.g., upgrading from SMS to telephone if no response is received). At the second time point, the enhanced intervention service level plan is pushed out by updating the intervention interaction method.

[0102] like Figure 8 As shown, in some embodiments, step S702 may include, but is not limited to, steps S801 to S803: Step S801: Extract features from the feedback behavior data to obtain feedback behavior features; wherein, the feedback behavior features include response timeliness features, interaction depth features, and conversion intention features.

[0103] Step S802: Update the fused user characteristics based on response timeliness characteristics, interaction depth characteristics, and conversion intention characteristics to obtain updated fused user characteristics; Step S803: Adjust the weights of the intervention interaction methods according to the updated and integrated user characteristics to obtain the updated intervention interaction methods.

[0104] In step S801 of some embodiments, specifically, the feedback behavior feature refers to the vector representation extracted from the feedback behavior data that characterizes the response behavior of the target insured user to the instant access service level plan.

[0105] Specifically, response timeliness is a vector representation that characterizes the time interval between when a target insured user receives the instant service level plan and when they generate a response.

[0106] Specifically, interaction depth features refer to vector representations that characterize the operational complexity and information acquisition depth of the target insured user in the feedback behavior.

[0107] Specifically, conversion intention features refer to the vector representation that characterizes the strength of the renewal intention shown by the target insured user in the feedback behavior.

[0108] Specifically, feedback behavioral data can be embedded to obtain a sequence of feedback behavioral vectors. Then, an RNN (Recurrent Neural Network) can be used to perform behavioral analysis on the behavioral vector sequence step by step to obtain time-series-based response timeliness features, interaction depth features, and conversion intention features.

[0109] In step S802 of some embodiments, specifically, updating the fused user features refers to the updated feature vector that reflects the latest behavioral status of the target insured user after integrating the response timeliness features, interaction depth features, and conversion intention features with the original fused user features.

[0110] Specifically, response timeliness features, interaction depth features, conversion intention features, and integrated user features can be combined to obtain updated integrated user features.

[0111] In step S803 of some embodiments, specifically, updating the intervention interaction method refers to the new interaction channel and form after optimizing and adjusting the original intervention interaction method based on feedback behavior data analysis.

[0112] Taking car insurance renewal as an example, for high-risk and high-value target policyholders, a weight table can be used to adjust the weight of the intervention interaction method. If the basic weights are SMS 0.3, regular phone 0.4, and dedicated phone 0.3, the weight of SMS can be reduced to 0.06 by multiplying it by 0.2 based on the response timeliness feature "no response" in the updated integrated user characteristics. The weight of regular phone can be increased to 0.48 by multiplying it by 1.2 based on the interaction depth feature "zero interaction". The weight of dedicated phone can be increased to 0.45 by multiplying it by 1.5 based on the conversion intention feature "no intention" and the user value level "high value level". At the same time, considering the urgency of the remaining 15 days of the policy, an additional timeliness weight of 0.1 is added to the telephone channels. Finally, the weights are calculated as SMS 0.06, regular phone 0.58, and dedicated phone 0.55. Regular phone with the highest weight is selected as the candidate, but according to the "high priority" requirement in the risk level intervention strategy, it is upgraded to dedicated phone. The updated intervention interaction method is determined to be "dedicated telephone follow-up".

[0113] Through steps S801 to S803, the system can perceive customers' feedback status to early intervention in real time during the critical window period for renewal decisions. For unresponsive customers, the interaction method is automatically upgraded to a high-intensity telephone channel. For high-value customers with high churn risk, dedicated service resources are prioritized. For low-value customers, the intensity of the interaction method upgrade is reduced. This achieves dual optimization of intervention resources in both channel and value dimensions, significantly improving the conversion efficiency of risk intervention service solutions.

[0114] Through steps S701 to S702, the system can accurately deliver preventative service solutions within limited interaction opportunities, taking into account the fundamental differences in decision-making based on the long policy cycle, low frequency of contact, significant churn, and dual dimensions of risk and value in insurance policies. This achieves dual optimization of intervention resources in both time and value dimensions, significantly improving the conversion efficiency of risk intervention service solutions.

[0115] As can be seen, in the above solution, target user data of the target insured user is obtained, and at least two of the target user features, including user behavior features, user policy features, and user interaction features, are extracted. At least two of these features are then fused to obtain fused user features. Based on these fused user features, the churn risk of the target insured user is assessed to obtain the degree of churn risk. Finally, a target risk intervention service plan is generated based on the degree of churn risk. In this invention, for scenarios with dynamic changes in insured user behavior and significant individual differences, at least two of the user behavior features, user policy features, and user interaction features can be extracted to capture dynamic user characteristics within the policy period. The degree of churn risk is then assessed based on the fused user features to generate a risk intervention service plan with individual differences. This effectively avoids the shortcomings of rigid business experience and uniform intervention strategies in traditional methods, significantly improving the matching effect between the risk intervention plan and the user's actual needs, thereby improving the conversion efficiency of the risk intervention service plan.

[0116] 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.

[0117] Please see Figure 9 This application also provides a risk intervention service plan generation device, which can implement the above-mentioned knowledge graph prediction method. The device includes: The user data acquisition module is used to acquire target user data of the target insured user. The feature extraction module is used to extract features from the target user data to obtain target user features; wherein, the target user features include at least two of the following: user behavior features on the pre-built insurance platform, user policy features, and user interaction features; The feature fusion module is used to fuse at least two of the user behavior features, user policy features, and user interaction features to obtain fused user features. The churn risk assessment module is used to assess the churn risk of target insured users based on the characteristics of integrated users, and to obtain the degree of churn risk of the target users. The intervention service plan generation module is used to generate a target risk intervention service plan for the target insured user based on the risk level of target user churn.

[0118] In one embodiment, the churn risk assessment module is specifically used for: Based on the integrated user characteristics, a preliminary churn assessment is conducted on the target insured users to obtain the initial user churn risk level; If the initial user churn risk level meets the predetermined churn risk conditions, then risk analysis is performed on the integrated user characteristics based on the initial user churn risk level to obtain the target churn risk characteristics; The risk of churn for target insured users is assessed based on the initial user churn risk level and the target churn risk characteristics to obtain the target user churn risk level.

[0119] In one embodiment, the churn risk assessment module is further specifically used for: The risk contribution of the merged user characteristics is calculated based on the initial user churn risk level to obtain the characteristic risk contribution degree; wherein, the characteristic risk contribution degree is used to assess the degree to which the merged user characteristics affect the churn risk of the target insured users; Based on the contribution of characteristic risks and predetermined contribution conditions, the characteristics of integrated users are screened to obtain the target churn risk characteristics.

[0120] In one embodiment, the churn risk assessment module is further specifically used for: Target feature risk contribution degree is used to screen out target churn risk features; Based on the risk contribution of the target characteristics and the initial user churn risk level, the target insured users are scored for churn risk, and a user churn risk score is obtained. Based on the characteristics of target churn risk, the value of target insured users is predicted to obtain user value data; among which, user value data is used to characterize the expected revenue that target insured users provide to the insurance platform; The risk level of target insured users is assessed based on user churn risk score and user value data to obtain the degree of target user churn risk; among which, the degree of target user process risk includes user churn risk level and user value level.

[0121] In one embodiment, the intervention service plan generation module is specifically used for: Feature extraction is performed on user churn risk level to obtain user churn risk level features, and feature extraction is performed on user value level to obtain user value level features; Intervention strategies are generated based on user churn risk level characteristics, user value level characteristics, and target churn risk characteristics to obtain risk level intervention strategies. Intervention service plans are generated based on a risk intervention generation model pre-trained according to risk level intervention strategy instructions, resulting in a target risk intervention service plan; among which, the target risk intervention service plan includes an immediate reach service level plan and an enhanced intervention service level plan.

[0122] In one embodiment, the risk intervention service plan generation device further includes: Get the first point in time before the policy expiry date of the target insured user, and push the instant outreach service level plan to the target insured user based on the first point in time and the intervention interaction method; After obtaining the feedback behavior data of the target insured user after the push, if the feedback behavior data indicates that there is no intention to renew the policy within the preset response period, then obtain the second time point of the target insured user before the policy expiration date, and update the intervention interaction method based on the feedback behavior data to obtain the updated intervention interaction method. Based on the updated intervention interaction method and the second time point, push the enhanced intervention service level plan to the target insured user; wherein, the first time point is earlier than the second time point.

[0123] In one embodiment, the risk intervention service plan generation device further includes: Feature extraction is performed on the feedback behavior data to obtain feedback behavior features; among which, feedback behavior features include response timeliness features, interaction depth features, and conversion intention features; The merged user characteristics are updated based on response timeliness, interaction depth, and conversion intention characteristics to obtain updated merged user characteristics; The intervention interaction methods are adjusted by weighting the updated and integrated user characteristics to obtain the updated intervention interaction methods.

[0124] Specific limitations regarding the risk intervention service plan generation device can be found in the limitations of the intelligent question-and-answer method described above, and will not be repeated here. Each module in the aforementioned risk intervention service plan generation 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, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0125] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 10 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 the computer program is executed by the processor, it implements the functions or steps of a risk intervention service scheme generation method on the server side.

[0126] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 11 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing 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 in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the client-side functions or steps of a risk intervention service scheme generation method.

[0127] 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 target user data for the target insured users; Feature extraction is performed on the target user data to obtain target user features; among which, target user features include at least two of the following: user behavior features on the pre-built insurance platform, user policy features, and user interaction features; At least two of the user behavior characteristics, user policy characteristics, and user interaction characteristics are fused to obtain fused user characteristics. Based on the integrated user characteristics, the churn risk assessment of the target insured users is conducted to obtain the degree of churn risk of the target users; Generate a target risk intervention service plan for the target insured users based on the risk level of target user churn.

[0128] 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 target user data for the target insured users; Feature extraction is performed on the target user data to obtain target user features; among which, target user features include at least two of the following: user behavior features on the pre-built insurance platform, user policy features, and user interaction features; At least two of the user behavior characteristics, user policy characteristics, and user interaction characteristics are fused to obtain fused user characteristics. Based on the integrated user characteristics, the churn risk assessment of the target insured users is conducted to obtain the degree of churn risk of the target users; Generate a target risk intervention service plan for the target insured users based on the risk level of target user churn.

[0129] 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.

[0130] 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.

[0131] 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.

[0132] It should be noted that any AI models, software tools, or components not belonging to this company appearing in the embodiments of this application are merely illustrative examples and do not represent actual use. All user personal information involved in the embodiments of this application has been authorized (with the knowledge and consent) by the relevant parties or has been fully authorized by all parties, and the executing entity may obtain it through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with relevant laws and regulations and do not violate public order and good morals.

[0133] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. 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 generating a risk intervention service plan, characterized in that, include: Obtain target user data for the target insured users; Feature extraction is performed on the target user data to obtain target user features; wherein, the target user features include at least two of the following: user behavior features on the pre-built insurance platform, user policy features, and user interaction features; At least two of the user behavior features, the user policy features, and the user interaction features are fused to obtain fused user features; Based on the integrated user characteristics, the churn risk assessment of the target insured users is performed to obtain the degree of churn risk of the target users; Based on the risk level of churn of the target users, a target risk intervention service plan is generated for the target insured users.

2. The method as described in claim 1, characterized in that, The step of assessing the churn risk of the target insured user based on the integrated user characteristics to obtain the degree of churn risk of the target user includes: Based on the integrated user characteristics, a preliminary churn assessment is conducted on the target insured users to obtain the initial user churn risk level; If the initial user churn risk level meets the predetermined churn risk conditions, then the integrated user characteristics are analyzed for risk based on the initial user churn risk level to obtain the target churn risk characteristics; Based on the initial user churn risk level and the target churn risk characteristics, the churn risk of the target insured user is assessed to obtain the target user churn risk level.

3. The method as described in claim 2, characterized in that, The step of performing risk analysis on the fused user characteristics based on the initial user churn risk level to obtain target churn risk characteristics includes: The risk contribution of the integrated user characteristics is calculated based on the initial user churn risk level to obtain the characteristic risk contribution degree; wherein, the characteristic risk contribution degree is used to assess the degree to which the integrated user characteristics affect the churn risk of the target insured user; Based on the risk contribution degree and predetermined contribution conditions, the fused user characteristics are filtered to obtain the target churn risk characteristics.

4. The method as described in claim 3, characterized in that, The step of assessing the churn risk of the target insured user based on the initial user churn risk level and the target churn risk characteristics to obtain the target user churn risk level includes: The target feature risk contribution degree of the target churn risk feature is selected from the feature risk contribution degree; Based on the risk contribution of the target feature and the initial user churn risk level, the target insured user is scored for churn risk, and a user churn risk score is obtained. Based on the target churn risk characteristics, the value of the target insured user is predicted to obtain user value data; wherein, the user value data is used to characterize the expected revenue provided by the target insured user to the insurance platform; The target insured user is assessed for value risk level based on the user churn risk score and the user value data to obtain the degree of target user churn risk.

5. The method as described in claim 2, characterized in that, The risk level of the target user process includes the user churn risk level and the user value level; The step of generating a target risk intervention service plan for the target insured user based on the target user churn risk level includes: Feature extraction is performed on the user churn risk level to obtain user churn risk level features, and feature extraction is performed on the user value level to obtain user value level features; Based on the user churn risk level characteristics, the user value level characteristics, and the target churn risk characteristics, an intervention strategy is generated to obtain a risk level intervention strategy. Based on the risk level intervention strategy indication, a pre-trained risk intervention generation model is used to generate an intervention service plan, thereby obtaining the target risk intervention service plan.

6. The method as described in claim 5, characterized in that, The target risk intervention service plan includes an immediate reach service level plan and an enhanced intervention service level plan; After generating the target risk intervention service plan by a risk intervention generation model pre-trained based on the risk level intervention strategy indication, the method further includes: Obtain the first time point before the policy expiration date of the target insured user, and push the instant outreach service level plan to the target insured user based on the first time point and the intervention interaction method; After the push notification, the system obtains the feedback behavior data of the target insured user. If the feedback behavior data indicates that the user does not have the intention to renew the policy within a preset response period, the system obtains the second time point of the target insured user before the policy expiration date and updates the intervention interaction method based on the feedback behavior data to obtain an updated intervention interaction method. Based on the updated intervention interaction method and the second time point, the system pushes the enhanced intervention service level plan to the target insured user. The first time point is earlier than the second time point.

7. The method as described in claim 6, characterized in that, The step of updating the intervention interaction method based on the feedback behavior data to obtain an updated intervention interaction method includes: Feature extraction is performed on the feedback behavior data to obtain feedback behavior features; wherein, the feedback behavior features include response timeliness features, interaction depth features, and conversion intention features; The merged user features are updated based on the response timeliness feature, the interaction depth feature, and the conversion intention feature to obtain the updated merged user features; The intervention interaction method is obtained by adjusting the weights of the updated and integrated user characteristics.

8. A risk intervention service plan generation device, characterized in that, include: The user data acquisition module is used to acquire target user data of the target insured user. The feature extraction module is used to extract features from the target user data to obtain target user features; wherein, the target user features include at least two of the following: user behavior features on the pre-built insurance platform, user policy features, and user interaction features; The feature fusion module is used to fuse at least two of the user behavior features, the user policy features, and the user interaction features to obtain fused user features. The churn risk assessment module is used to assess the churn risk of the target insured user based on the integrated user characteristics, and to obtain the degree of churn risk of the target user. The intervention service plan generation module is used to generate a target risk intervention service plan for the target insured user based on the target user's churn risk level.

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, When the processor executes the computer program, it implements the steps of the risk intervention service scheme generation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the risk intervention service scheme generation method as described in any one of claims 1 to 7.

Images