Business processing method and device, electronic equipment and medium

Abstract

The embodiment of the invention provides a business processing method and device, electronic equipment and a medium, belongs to the technical field of financial science and technology, and can be applied to application scenes of financial insurance and medical health. The method comprises the steps of obtaining market economic data of a target service and feature data of a target user in response to a handling request of the target user for the target service; generating at least two feature combinations according to each of the market economic data and the feature data; and inputting the feature combination into a service processing model, determining a service index corresponding to each sub-model through each sub-model of the service processing model based on the feature combination corresponding to each sub-model in the feature combination, and generating a service processing scheme corresponding to the target service and the target user based on the service index corresponding to each sub-model. According to the embodiment of the invention, the market economic data and the multi-dimensional user features are dynamically integrated, the multi-dimensional feature combination is constructed, and the sub-models are used for cooperative processing, so that dynamic optimization and accurate decision-making of the business processing scheme are realized.

Description

Technical Field

[0001] This application relates to the field of financial technology, and more particularly to a business processing method, apparatus, electronic device, and medium. Background Technology

[0002] Currently, insurance companies in the industry suffer from severe homogenization in product design, lacking innovation and leading to decreased product competitiveness and compressed profit margins. Furthermore, they exhibit significant weaknesses in risk identification and assessment, particularly in their insufficient ability to predict the impact of extreme risk events on liability costs, resulting in considerable uncertainty in risk management and capital allocation decisions. In addition, insurance companies generally lack intelligent and digital liability-side cost management tools, relying excessively on manual experience and judgment. This makes it difficult to efficiently process massive amounts of data and conduct complex scenario analyses, resulting in low cost management efficiency, delayed decision-making, and severely restricting their market competitiveness and risk control capabilities.

[0003] Existing business processing methods rely on fixed risk assessment models and predetermined interest rates, and process business based on historical market data. This leads to unsatisfactory business processing results, rigid pricing, and an inability to adapt to complex and ever-changing market environments. Consequently, cost management lags behind market changes. Therefore, improving the accuracy and efficiency of business processing has become an urgent technical problem to be solved. Summary of the Invention

[0004] The main objective of this application is to propose a business processing method, apparatus, electronic device, and medium, which aims to solve the problems of existing technologies relying on fixed templates, making it difficult to dynamically respond to market and customer risk preferences, resulting in poor cost management and post-decision issues.

[0005] To achieve the above objectives, a first aspect of this application proposes a service processing method applied to an electronic device, the electronic device including a service processing model, the service processing model including at least one sub-model, the method including:

[0006] In response to a target user's request to process a target service, the system acquires market economic data of the target service and characteristic data of the target user; the characteristic data includes user behavior data, historical business data, and risk preference data.

[0007] Based on the market economy data and the data in the feature data, generate at least two feature combinations;

[0008] The at least two feature combinations are input into the business processing model. Based on the feature combinations corresponding to each sub-model in the at least two feature combinations, the business indicators corresponding to each sub-model are determined by each sub-model. Based on the business indicators corresponding to each sub-model, a business processing plan corresponding to the target business and the target user is generated.

[0009] In some embodiments, when the target business is insurance policy business, the business processing model includes a policy cost optimization model, a surrender cost prediction model, and a reserve cost optimization model. The process involves inputting the at least two feature combinations into the business processing model, and then determining the business indicators corresponding to each sub-model based on the feature combinations corresponding to the at least two feature combinations. This includes:

[0010] The first feature combination is input into the policy cost optimization model to obtain the policy cost index output by the policy cost optimization model; the first feature combination is generated based on the market economic data, the target user's user behavior data, and risk preference data.

[0011] The second feature combination is input into the surrender cost prediction model to obtain the surrender cost index output by the surrender cost prediction model; the second feature combination is generated based on the target user's user behavior data, historical business data, and risk preference data.

[0012] The third feature combination is input into the reserve cost optimization model to obtain the reserve cost index output by the reserve cost optimization model; the third feature combination is generated based on market economic data and user behavior data of the target user.

[0013] In some embodiments, before inputting the at least two feature combinations into the business processing model, and determining the business metrics corresponding to each sub-model based on the feature combinations corresponding to each sub-model from the at least two feature combinations, the method further includes:

[0014] Obtain historical insurance policy dataset;

[0015] Based on the random forest model, construct an initial policy cost optimization model;

[0016] A gradient boosting tree with temporal attention mechanism is used to construct an initial surrender cost prediction model;

[0017] Based on the pre-defined deep learning model, construct an initial reserve cost optimization model;

[0018] Based on the historical insurance policy dataset, the initial policy cost optimization model, the initial surrender cost prediction model, and the initial reserve cost optimization model are trained to obtain the policy cost optimization model, the surrender cost prediction model, and the reserve cost optimization model.

[0019] In some embodiments, generating a business processing plan corresponding to the target business and the target user based on the business metrics corresponding to each sub-model includes:

[0020] The comprehensive business indicators for the target business are calculated based on the policy cost indicator, the surrender cost indicator, and the reserve cost indicator.

[0021] If the surrender cost index is less than or equal to a preset surrender cost index threshold, a renewal discount scheme for the target business is generated.

[0022] Based on the renewal discount scheme, a business processing scheme for the target business is generated according to the comprehensive business indicators of the target business.

[0023] In some embodiments, before acquiring the market economic data of the target business and the characteristic data of the target user, the method further includes:

[0024] Acquire the initial macroeconomic data, initial market trend data, initial user behavior data, initial historical business data, and initial risk preference data for the target business;

[0025] The initial macroeconomic data, the initial user behavior data, and the initial risk preference data are subjected to anomaly processing to obtain the macroeconomic data, user behavior data, and risk preference data; the anomaly processing includes at least missing value processing and outlier processing.

[0026] The initial market trend data is smoothed using a sliding window method to obtain the market trend data.

[0027] The initial historical business data is processed using a box plot or the Raida criterion to obtain the historical business data.

[0028] In some embodiments, the market economic data includes macroeconomic data and market trend data, and the step of generating at least two feature combinations based on the market economic data and each of the feature data includes:

[0029] Data features are extracted from the macroeconomic data, market trend data, user behavior data, historical business data, and risk preference data to obtain macroeconomic features, market trend features, user behavior features, historical business features, and risk preference features, respectively.

[0030] The user behavior characteristics, the market trend characteristics, and the risk preference characteristics are combined to obtain the first feature combination;

[0031] The user behavior characteristics, historical business characteristics, and risk preference characteristics are combined to obtain the second feature combination;

[0032] By combining the user behavior characteristics, the macroeconomic characteristics, and the market trend characteristics, a third characteristic combination is obtained;

[0033] The at least two feature combinations include at least the first feature combination, the second feature combination, and the third feature combination.

[0034] In some embodiments, after generating the business processing plan corresponding to the target business and the target user based on the business metrics corresponding to each sub-model, the method includes:

[0035] Under the condition that the target conditions are met, obtain the market interest rate fluctuation signal of the target business, the competitor pricing strategy change event, and the newly added historical business dataset;

[0036] Based on the market interest rate fluctuation signal, the competitor pricing strategy change event, and the newly added historical business dataset, the parameters of the policy cost optimization model, the surrender cost prediction model, and the reserve cost optimization model are updated.

[0037] The target conditions include: the comprehensive business indicator generated based on each of the business indicators is greater than or equal to the preset business indicator threshold.

[0038] To achieve the above objectives, a second aspect of this application provides a service processing apparatus, the apparatus comprising:

[0039] The response module is used to respond to a target user's request to process a target service, and to obtain market economic data of the target service and characteristic data of the target user; the characteristic data includes user behavior data, historical business data and risk preference data.

[0040] The generation module is used to generate at least two feature combinations based on the market economy data and each data in the feature data;

[0041] The processing module is used to input the at least two feature combinations into the business processing model, determine the business indicators corresponding to each sub-model based on the feature combinations corresponding to each sub-model among the at least two feature combinations, and generate the business processing scheme corresponding to the target business and the target user based on the business indicators corresponding to each sub-model.

[0042] To achieve the above objectives, a third aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect.

[0043] To achieve the above objectives, a fourth aspect of the present application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect.

[0044] The business processing method, apparatus, electronic device, and medium proposed in this application respond to a target user's request for processing a target business by acquiring market economic data of the target business and characteristic data of the target user; generating at least two feature combinations based on each data in the market economic data and characteristic data; inputting the feature combinations into a business processing model; determining the business indicators corresponding to each sub-model based on the feature combinations corresponding to each sub-model in the business processing model; and generating a business processing scheme corresponding to the target business and the target user based on the business indicators corresponding to each sub-model. This achieves dynamic optimization and accurate decision-making of the business processing scheme by dynamically integrating market economic data and multi-dimensional user characteristics, constructing multi-dimensional feature combinations, and utilizing sub-models for collaborative processing. Attached Figure Description

[0045] Figure 1 This is a flowchart of the business processing method provided in the embodiments of this application;

[0046] Figure 2 yes Figure 1 A flowchart of step S103 in the process;

[0047] Figure 3 yes Figure 1 Another flowchart of step S103 in the process;

[0048] Figure 4 yes Figure 1 Another flowchart of step S103 in the process;

[0049] Figure 5 yes Figure 4 The flowchart of step S101 in the text;

[0050] Figure 6 yes Figure 2 The flowchart of step S102 in the document;

[0051] Figure 7 yes Figure 6 Another flowchart of step S103 in the process;

[0052] Figure 8 This is a schematic diagram of the structure of the business processing device provided in the embodiments of this application;

[0053] Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0055] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0057] Currently, insurance companies in the industry suffer from severe homogenization in product design, lacking innovation and leading to decreased product competitiveness and compressed profit margins. Furthermore, they exhibit significant weaknesses in risk identification and assessment, particularly in their insufficient ability to predict the impact of extreme risk events on liability costs, resulting in considerable uncertainty in risk management and capital allocation decisions. In addition, insurance companies generally lack intelligent and digital liability-side cost management tools, relying excessively on manual experience and judgment. This makes it difficult to efficiently process massive amounts of data and conduct complex scenario analyses, resulting in low cost management efficiency, delayed decision-making, and severely restricting their market competitiveness and risk control capabilities.

[0058] Existing business processing methods rely on fixed risk assessment models and predetermined interest rates, and process business based on historical market data. This leads to unsatisfactory business processing results, rigid pricing, and an inability to adapt to complex and ever-changing market environments. Consequently, cost management lags behind market changes. Therefore, improving the accuracy and efficiency of business processing has become an urgent technical problem to be solved.

[0059] Based on this, embodiments of this application provide a business processing method, apparatus, electronic device, and medium, aiming to improve the solution to the problems of existing technologies relying on fixed templates, making it difficult to dynamically respond to market and customer risk preferences, resulting in poor cost management and post-decision issues.

[0060] The business processing methods, apparatus, electronic devices and media provided in the embodiments of this application are specifically described through the following embodiments. First, the business processing methods in the embodiments of this application are described.

[0061] 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.

[0062] 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.

[0063] The business processing method provided in this application relates to the field of financial technology and can be applied to application scenarios such as financial insurance and healthcare. The business processing method provided in this application can also be applied to electronic devices. In some embodiments, the electronic device may be a smartphone, tablet computer, laptop computer, desktop computer, etc., but is not limited to these forms.

[0064] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0065] It should be noted that in all specific embodiments of this application, when processing data related to user identity or characteristics, such as user information, user behavior data, user historical data, and user location information, user permission or consent is obtained first. Furthermore, the collection, use, and processing of this data comply with relevant laws, regulations, and standards. In addition, when embodiments of this application require access to sensitive personal information of users, separate permission or consent from the user is obtained through pop-ups or redirection to confirmation pages. Only after obtaining the user's separate permission or consent is the necessary user-related data required for the proper functioning of these embodiments acquired.

[0066] Figure 1 This is an optional flowchart of the business processing method provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S101 to S103.

[0067] Step S101: In response to the target user's request to process the target service, obtain the market economic data of the target service and the characteristic data of the target user; the characteristic data includes user behavior data, historical business data and risk preference data.

[0068] Step S102: Generate at least two feature combinations based on the data in the market economy data and feature data;

[0069] Step S103: Input at least two feature combinations into the business processing model. Based on the feature combinations corresponding to each sub-model in the at least two feature combinations, determine the business indicators corresponding to each sub-model. Based on the business indicators corresponding to each sub-model, generate a business processing plan corresponding to the target business and the target user.

[0070] In this step, the electronic device includes a business processing model, which comprises at least one sub-model. When a target user requests to process a target service, the system first acquires market economic data for the target service and characteristic data for the target user. This characteristic data includes user behavior data, historical business data, and risk preference data. This characteristic data covers multiple aspects of the market environment and the user's personal circumstances, providing a comprehensive information foundation for subsequent processing.

[0071] Specifically, the business processing model refers to the algorithm model used to process the target business. It can be implemented by an integrated model containing multiple sub-models. By processing different feature combinations of different dimensions through different sub-models, the problem that traditional fixed models cannot adapt to complex market environments can be solved. A sub-model refers to an independent model in the business processing model that is specifically designed to process specific business indicators. It can be implemented by machine learning models or deep learning models. By dividing the work among sub-models to process different feature combinations, the accuracy of business indicator prediction can be improved.

[0072] Furthermore, based on the various data in the acquired market economic data and feature data, at least two feature combinations are generated. The generation process of feature combinations involves the selection and combination of different types of data, aiming to extract the most relevant and valuable information, thereby achieving the goal of improving the completeness of user profiles through multi-dimensional data fusion to support accurate decision-making.

[0073] Specifically, market economy data refers to macroeconomic indicators and industry trend data that reflect market dynamics, including interest rate fluctuations, changes in market supply and demand, and policy and regulatory information. By acquiring and integrating such data in real time, the problem of decision-making lag caused by the reliance on historical data in traditional methods can be solved.

[0074] Furthermore, at least two feature combinations generated are input into the business processing model. Each sub-model in the business processing model processes its corresponding feature combination. Based on the input feature combination, each sub-model determines the corresponding business indicators, so as to carry out specialized analysis and prediction for different types of data and business characteristics through modular processing.

[0075] Furthermore, based on the business indicators determined by each sub-model, business processing solutions corresponding to target businesses and target users are generated. The preferred solution can be generated through weighted calculation or rule engine. The output results of each sub-model are comprehensively considered to form a comprehensive business processing solution. Ultimately, through the acquisition of multi-dimensional data, the generation of feature combinations, multi-model collaborative processing, and comprehensive decision-making, refined and personalized management of business processing is achieved, thus realizing the accuracy and personalization of business processing.

[0076] For example, in an insurance business scenario, the electronic device serves as the insurance company's business processing system. The business processing model includes three sub-models: a policy pricing model, a claims prediction model, and a customer churn risk model. When customer A applies to purchase a life insurance policy, the system first acquires macroeconomic data such as the current inflation rate and GDP growth rate, as well as market trend data for the insurance industry, serving as market economic data. Simultaneously, it acquires user behavior data such as customer A's age, occupation, income level, and past insurance records, historical business data such as policy purchases and claims history over the past five years, and customer A's risk attitude assessment results, serving as risk preference data.

[0077] Furthermore, based on the acquired market economic data, user behavior data, historical business data, and risk preference data, three feature combinations are generated: the first feature combination includes macroeconomic indicators, market trends, and customer personal attributes; the second feature combination includes the customer's historical business data and risk preferences; and the third feature combination includes the customer's behavioral data and market trend data.

[0078] Furthermore, the first feature combination is input into the policy pricing model, the second feature combination into the claims prediction model, and the third feature combination into the customer churn risk model to obtain the output results of each model: policy pricing recommendations, expected claims ratio, and customer churn risk score. Then, by combining these three indicators, a personalized insurance plan for customer A is generated, including recommended insurance products, premium pricing, and service strategies.

[0079] Steps S101 to S102, as illustrated in this embodiment, involve responding to a target user's request for processing a target service by acquiring market economic data of the target service and characteristic data of the target user; generating at least two feature combinations based on each data point in the market economic data and characteristic data; inputting the feature combinations into a business processing model; determining business indicators corresponding to each sub-model based on the feature combinations corresponding to each sub-model in the feature combinations; and generating a business processing solution corresponding to the target service and the target user based on the business indicators corresponding to each sub-model. This approach dynamically integrates market economic data and multi-dimensional user characteristics to construct multi-dimensional feature combinations and utilizes sub-models for collaborative processing, thereby achieving dynamic optimization and accurate decision-making of the business processing solution.

[0080] Please see Figure 2 In some embodiments, in step S103, at least two feature combinations are input into the business processing model. Each sub-model of the business processing model determines its corresponding business indicators based on the feature combinations corresponding to each of the at least two feature combinations. This may include, but is not limited to, steps S201 to S203.

[0081] Step S201: Input the first feature combination into the policy cost optimization model to obtain the policy cost index output by the policy cost optimization model; the first feature combination is generated based on market economic data, user behavior data of target users, and risk preference data.

[0082] Step S202: Input the second feature combination into the surrender cost prediction model to obtain the surrender cost index output by the surrender cost prediction model; the second feature combination is generated based on the target user's user behavior data, historical business data and risk preference data;

[0083] Step S203: Input the third feature combination into the reserve cost optimization model to obtain the reserve cost index output by the reserve cost optimization model; the third feature combination is generated based on market economic data and user behavior data of target users.

[0084] In this step, when the target business is insurance policy business, the business processing model includes a policy cost optimization model, a surrender cost prediction model, and a reserve cost optimization model. At least two feature combinations are input into the business processing model. Based on the feature combinations corresponding to each sub-model from the at least two feature combinations, the business indicators corresponding to each sub-model are determined by the sub-models of the business processing model.

[0085] Specifically, the first feature combination is input into the policy cost optimization model to obtain the policy cost index output by the model. The first feature combination is generated based on market economic data, target user behavior data, and risk preference data. For example, market economic data includes macroeconomic indicators such as GDP growth rate and inflation rate; user behavior data includes policy purchase frequency and claims records; and risk preference data includes users' investment preferences and risk tolerance. These data are processed using feature engineering to form the first feature combination.

[0086] Furthermore, the second feature combination is input into the surrender cost prediction model to obtain the surrender cost index output by the model. The second feature combination is generated based on the target user's user behavior data, historical business data, and risk preference data. Among them, historical business data may include information such as the user's past policy renewal rate and claims rate.

[0087] Furthermore, the third feature combination is input into the reserve cost optimization model to obtain the reserve cost index output by the model. The third feature combination is generated based on market economic data and user behavior data of the target users. In addition, the reserve cost optimization model can consider factors such as changes in market interest rates and user claim probabilities to optimize the reserve size.

[0088] Furthermore, by inputting different feature combinations into specialized sub-models, policy costs, surrender costs, and reserve costs can be assessed in a targeted manner. This allows for a more accurate grasp of business risks and cost structures, improving the model's flexibility and scalability, and enabling it to adapt to different types of insurance business scenarios. Simultaneously, by fully utilizing market economic data and user characteristic data, the model's predictions are more closely aligned with reality, helping insurance companies formulate more precise pricing strategies and risk management measures. This not only improves the accuracy and efficiency of business processing but also enhances the insurance companies' ability to respond to market changes.

[0089] Preferably, the policy cost optimization model generates policy cost indicators by integrating macroeconomic fluctuations, user behavior characteristics, and risk preference data. For example, it combines market interest rate changes with user insurance preferences to calculate premium adjustment thresholds. The surrender cost prediction model constructs a surrender probability prediction function based on users' historical policy operation records, current behavioral patterns, and risk tolerance. Specifically, it uses time series analysis to quantify the impact of user behavior on surrender costs. The reserve cost optimization model combines market trend prediction data and user activity indicators, and generates reserve provision ratios through a dynamic weight allocation algorithm. For example, it automatically increases the reserve coverage ratio when market volatility exceeds a preset threshold.

[0090] For example, in the context of insurance policy business, the first feature combination includes macroeconomic indicators, user operation logs, and risk assessment results. After being input into the policy cost optimization model, the model calculates the optimal premium pricing range through regression analysis. The second feature combination integrates users' historical claims records, recent interaction behavior, and risk preference levels. The surrender cost prediction model predicts the future surrender rate based on the gradient boosting tree algorithm and outputs the corresponding estimated financial loss. The third feature combination integrates market trend index and user policy activity data. The reserve cost optimization model uses Monte Carlo simulation to generate reserve requirements under different market scenarios to ensure that the capital adequacy ratio meets regulatory requirements.

[0091] By separating the calculation logic of different cost types and matching them with exclusive feature combinations, each sub-model can process core business indicators in parallel. Among them, the surrender cost prediction model corrects the prediction results in real time after the user behavior data is updated, and the reserve cost optimization model dynamically adjusts the optimization cycle according to market data fluctuations, thereby achieving precise collaboration in multi-dimensional cost management.

[0092] Please see Figure 3 In some embodiments, before inputting at least two feature combinations into the business processing model in step S103, and determining the business indicators corresponding to each sub-model based on the feature combinations corresponding to each sub-model from the at least two feature combinations, the method may also include, but is not limited to, steps S301 to S305:

[0093] Step S301: Obtain the historical insurance policy dataset;

[0094] Step S302: Construct an initial policy cost optimization model based on the random forest model;

[0095] Step S303: Use a gradient boosting tree with temporal attention mechanism to construct an initial surrender cost prediction model;

[0096] Step S304: Construct an initial reserve cost optimization model based on a preset deep learning model;

[0097] Step S305: Train the initial policy cost optimization model, the initial surrender cost prediction model, and the initial reserve cost optimization model based on the historical insurance policy dataset to obtain the policy cost optimization model, the surrender cost prediction model, and the reserve cost optimization model.

[0098] In this step, the historical insurance policy dataset is divided into a training set and a validation set after data cleaning and feature engineering. The historical insurance policy dataset contains multi-dimensional time-series business data, covering policy records under different market cycles.

[0099] Furthermore, the initial policy cost optimization model specifically uses parallel training of random forests to quickly generate multiple tree structures, and a voting mechanism to select the optimal cost prediction results. By integrating the prediction results of multiple decision trees, the random forest model effectively reduces the risk of overfitting and is suitable for processing high-dimensional feature data in policy cost optimization.

[0100] Furthermore, the initial surrender cost prediction model introduces an attention mechanism into the gradient boosting tree framework to dynamically adjust the impact weights of surrender events within the historical time window, thereby improving the model's ability to respond to sudden surrender events. In this way, the gradient boosting tree with a temporal attention mechanism can capture the time dependence of historical surrender behavior and enhance the model's ability to predict surrender trends.

[0101] Furthermore, by using convolutional and fully connected layers in the deep learning network, the nonlinear mapping relationship between market trends and user profiles is automatically learned to construct an initial reserve cost optimization model. The preset deep learning model adopts a multi-layer nonlinear transformation structure, which can extract high-order correlation features from macroeconomic fluctuations and user behavior data to optimize the calculation accuracy of reserve costs.

[0102] Furthermore, based on historical insurance policy datasets, the initial policy cost optimization model, the initial surrender cost prediction model, and the initial reserve cost optimization model were trained separately. During training, each sub-model iteratively updated its parameters using the backpropagation algorithm, and each was designed specifically for the characteristics of different business indicators. Historical datasets were used for cross-validation during model training to ensure the generalization performance of each sub-model on specific tasks, until the loss function converged to a preset threshold. Thus, the trained sub-models can output high-precision policy cost indicators, surrender cost indicators, and reserve cost indicators, providing reliable data support for dynamically generating business processing solutions.

[0103] Please see Figure 4 In some embodiments, the process of generating a business processing scheme for the target business and the target user based on the business metrics corresponding to each sub-model in step S103 may include, but is not limited to, steps S401 to S403:

[0104] Step S401: Calculate the comprehensive business indicators for the target business based on the policy cost indicator, surrender cost indicator, and reserve cost indicator.

[0105] Step S402: If the surrender cost index is less than or equal to the preset surrender cost index threshold, generate a renewal discount plan for the target business.

[0106] Step S403: Based on the renewal discount scheme, generate a business processing plan for the target business according to the comprehensive business indicators of the target business.

[0107] In this step, when the output of the surrender cost prediction model is lower than the preset threshold, which can be determined by the relationship between historical surrender rates and profit margins, a renewal discount calculation instruction is triggered. By combining the output results of the policy cost optimization model and the reserve cost optimization model, a multi-objective optimization algorithm is then used to calculate the comprehensive business indicators.

[0108] Specifically, the calculation of comprehensive business indicators can be achieved through weighted summation or nonlinear function fusion, and the weights of different business indicators are dynamically adjusted based on historical data. For example, the policy cost indicator can be multiplied by a weighting factor of 0.6, and the reserve cost indicator can be multiplied by a weighting factor of 0.4 for linear combination.

[0109] Furthermore, based on the calculation results, a pre-defined discount rule library is invoked to match the corresponding rate discount ratio, generating a renewal discount plan that includes options for new premium rates, coverage adjustments, and payment periods. Finally, the renewal discount plan is matched and validated against user risk preference data, and options that do not meet the user's risk tolerance are eliminated before the final business processing plan is output.

[0110] This allows for a comprehensive assessment of the target business's status based on multiple business indicators, and the generation of targeted renewal incentive plans and business processing solutions based on the assessment results. This approach considers multiple factors and improves the accuracy and flexibility of business processing. Simultaneously, by setting thresholds for surrender costs, surrender risks can be effectively controlled, protecting the interests of the insurance company. Furthermore, adjusting business processing solutions based on comprehensive business indicators enables differentiated handling of different situations, improving customer satisfaction and business efficiency.

[0111] Please see Figure 5 In some embodiments, before obtaining the market economic data of the target business and the characteristic data of the target user in step S101, the method may also include, but is not limited to, steps S501 to S504:

[0112] Step S501: Obtain initial macroeconomic data, initial market trend data, initial user behavior data, initial historical business data, and initial risk appetite data for the target business;

[0113] Step S502: Perform anomaly processing on the initial macroeconomic data, initial user behavior data, and initial risk preference data to obtain macroeconomic data, user behavior data, and risk preference data; anomaly processing includes at least missing value handling and outlier handling;

[0114] Step S503: The initial market trend data is smoothed using the sliding window method to obtain the market trend data.

[0115] Step S504: Using a box plot or the Laida criterion, process the initial historical business data to obtain historical business data.

[0116] In this step, initial macroeconomic data, initial market trend data, initial user behavior data, initial historical business data, and initial risk appetite data for the target business are obtained. Initial macroeconomic data includes GDP growth rate, CPI index, unemployment rate, etc.; initial market trend data includes insurance industry growth rate, competitor pricing strategies, etc.; initial user behavior data includes user browsing history, purchase frequency, etc.; initial historical business data includes historical policy information, claims records, etc.; and initial risk appetite data includes user risk assessment questionnaire results, etc.

[0117] Furthermore, the initial macroeconomic data, initial user behavior data, and initial risk preference data undergo anomaly processing to obtain macroeconomic data, user behavior data, and risk preference data, respectively. Anomaly processing includes missing value handling and outlier handling, specifically by filling in missing values ​​or removing outliers to eliminate incomplete or erroneous information in the data.

[0118] Furthermore, the sliding window method is used to smooth the initial market trend data to obtain market trend data, and then the moving average of the market trend data is calculated using a fixed time window to suppress short-term fluctuation noise.

[0119] Furthermore, by employing box plots or the Raida criterion, the initial historical business data is processed to obtain historical business data, which effectively identifies and processes outlier data points, improves the representativeness and reliability of the data, lays the foundation for subsequent feature extraction and model training, and helps improve the accuracy and effectiveness of the final business processing solution. Specifically, box plots identify outliers in historical business data based on quartiles, while the Raida criterion filters outlier data through a threshold of three standard deviations.

[0120] Please see Figure 6 In some embodiments, generating at least two feature combinations in step S102 based on the market economy data and feature data may include, but is not limited to, steps S601 to S604:

[0121] Step S601: Extract data features from macroeconomic data, market trend data, user behavior data, historical business data, and risk preference data respectively to obtain macroeconomic features, market trend features, user behavior features, historical business features, and risk preference features.

[0122] Step S602: Combine user behavior characteristics, market trend characteristics, and risk preference characteristics to obtain the first feature combination;

[0123] Step S603: Combine user behavior characteristics, historical business characteristics, and risk preference characteristics to obtain the second feature combination;

[0124] Step S604: Combine user behavior characteristics, macroeconomic characteristics, and market trend characteristics to obtain the third characteristic combination;

[0125] Among them, at least two feature combinations include at least a first feature combination, a second feature combination, and a third feature combination.

[0126] In this step, market economic data includes macroeconomic data and market trend data. The steps for generating at least two feature combinations based on the market economic data and feature data include:

[0127] First, data features are extracted from macroeconomic data, market trend data, user behavior data, historical business data, and risk preference data to obtain macroeconomic features, market trend features, user behavior features, historical business features, and risk preference features.

[0128] Specifically, when extracting data features from macroeconomic data, market trend data, user behavior data, historical business data, and risk preference data, dimensionality reduction methods such as principal component analysis and factor analysis can be used to extract the most representative features. For example, macroeconomic features are extracted by analyzing economic cycle indicators and industry policy data; market trend features are generated by calculating volatility and trend slope based on time-series data after sliding window smoothing; user behavior features cover user operation frequency, page dwell time, and interaction path clustering results; historical business features are obtained by statistically analyzing users' past policy types, claims records, and renewal periods; and risk preference features are determined by the results of user investment choice questionnaires and the output of risk tolerance assessment models.

[0129] Furthermore, combining user behavior characteristics, market trend characteristics, and risk preference characteristics yields the first characteristic combination; combining user behavior characteristics, historical business characteristics, and risk preference characteristics yields the second characteristic combination; and combining user behavior characteristics, macroeconomic characteristics, and market trend characteristics yields the third characteristic combination.

[0130] Specifically, in the first feature combination, user behavior features and market trend features are linked to policy pricing dynamics, and risk preference features are used to adjust cost sensitivity; in the second feature combination, historical business features and user behavior features are combined to identify surrender behavior patterns, and risk preference features help predict surrender decision thresholds; in the third feature combination, macroeconomic features and market trend features jointly influence reserve provision strategies, and user behavior features are used to correct for individual differences in market signal responses.

[0131] Furthermore, in the combination process, the first feature combination weights and concatenates user behavior activity level, market trend residual component, and risk tolerance score. When input into the policy cost optimization model, the model can adjust the cost weights based on the real-time matching of market fluctuations and user risk preferences. The second feature combination performs cross-feature calculation on the distribution of users' historical policy types, frequency of surrender operations, and risk questionnaire results, enabling the surrender cost prediction model to capture the correlation between individual historical behavior and decision-making tendencies. The third feature combination generates a composite index by multiplying macroeconomic factor loadings and market trend cycle components. Combined with user behavior clustering results, the reserve cost optimization model can distinguish group differences under the influence of the macro environment.

[0132] Ultimately, the dimensionality and information density of the input data for each sub-model are optimized, avoiding noise interference from a single data source. Simultaneously, the business logic correlation between features is ensured, thereby improving the calculation accuracy of comprehensive business indicators and enhancing the accuracy and adaptability of business processing. This makes insurance product pricing and risk assessment more aligned with actual conditions. Furthermore, it allows for flexible responses to market changes and timely adjustments to feature combinations, thereby enhancing the market competitiveness and risk control capabilities of insurance companies.

[0133] Please see Figure 7 In some embodiments, after generating the business processing scheme corresponding to the target business and the target user based on the business indicators corresponding to each sub-model in step S103, the method may also include, but is not limited to, steps S701 to S702:

[0134] Step S701: If the target conditions are met, obtain the market interest rate fluctuation signal of the target business, the competitor pricing strategy change event, and the newly added historical business dataset.

[0135] Step S702: Based on market interest rate fluctuation signals, competitor pricing strategy change events, and newly added historical business datasets, update the parameters of the policy cost optimization model, surrender cost prediction model, and reserve cost optimization model.

[0136] The target conditions include: the comprehensive business indicator generated based on each business indicator is greater than or equal to the preset business indicator threshold.

[0137] In this step, under the condition that the target conditions are met, the market interest rate fluctuation signal of the target business, the competitor pricing strategy change event, and the newly added historical business dataset are obtained. The target conditions include that the comprehensive business indicator generated based on each business indicator is greater than or equal to the preset business indicator threshold. The market interest rate fluctuation signal is obtained by collecting benchmark interest rate data of the financial market in real time and calculating the volatility. The signal is generated when the volatility exceeds the preset threshold.

[0138] Specifically, by monitoring market interest rate trends, a market interest rate fluctuation signal is generated when the detected interest rate fluctuation exceeds a preset threshold. For example, a market interest rate fluctuation signal is triggered when the benchmark interest rate for bank deposits changes by more than 0.25 percentage points.

[0139] Furthermore, by tracking competitors' product pricing strategies in real time, significant changes in the pricing of competitor insurance products can be recorded as a competitor pricing strategy change event. For example, a competitor's pricing strategy change event is triggered when the premium for a similar insurance product is reduced by more than 10%. In addition, historical business data such as newly added policy data and claims data can be continuously collected to form a new historical business dataset. For example, newly added policy underwriting information and claims cases can be summarized monthly.

[0140] Furthermore, based on the acquired market interest rate fluctuation signals, competitor pricing strategy change events, and newly added historical business datasets, the parameters of the policy cost optimization model, surrender cost prediction model, and reserve cost optimization model are updated.

[0141] Specifically, newly added market interest rate data and competitor pricing data are input into the policy cost optimization model to retrain the model parameters to adapt to the latest market environment. Newly added policy data and claims data are input into the surrender cost prediction model and reserve cost optimization model to update the model parameters, improve prediction accuracy, and enable the model to learn the latest business models and risk characteristics, thereby continuously optimizing business processing solutions.

[0142] Ultimately, this enables dynamic updates to the business processing model, improving its sensitivity and adaptability to market changes. Furthermore, insurance companies can adjust their business strategies promptly, enhancing their risk management capabilities and market competitiveness. In addition, by incorporating external factors such as market interest rates and competitor pricing, the model's input dimensions are enriched, improving its predictive accuracy through continuous updates of historical business data.

[0143] In some embodiments, by integrating multi-source data and constructing and optimizing dynamic pricing models, precise optimization of insurance product design and pricing strategies can be achieved, thereby reducing costs on the liability side. The technical solution of this embodiment is described in detail below:

[0144] 1. Data Acquisition: Collect multi-source aggregated data (not from a single user) through public databases, market research data, and internal corporate databases. This includes historical policy data, macroeconomic indicators, customer behavior data, market trend data, and risk preference data. Historical policy data includes policy type, sum assured, and claims records. Macroeconomic indicators include GDP, interest rates, and inflation rates. Customer behavior data includes policy purchase / cancellation records and interaction frequency. Market trend data includes industry premium growth rates and competitor pricing. Risk preference data includes customer risk tolerance questionnaire results.

[0145] For example, historical policy data: obtaining users' historical policy information from the company's internal database, including policy type (such as whole life insurance, annuity insurance), sum insured, premium payment method (lump sum / periodic payment), claims records, etc.; user behavior data: obtaining users' policy purchase / cancellation records, online interaction frequency (such as number of app logins, number of customer service inquiries), policy renewal status, etc. through the company's internal business systems; risk preference data: obtaining data from users' risk tolerance questionnaires, including users' risk tolerance, investment preferences, etc.; macroeconomic indicators: obtaining GDP growth rate, market interest rates (such as 10-year treasury bond yield), inflation rate, etc. through public databases; market trend data: obtaining market premium growth rate, competitor pricing strategies, industry surrender rate, etc. from industry databases.

[0146] 2. Data Processing:

[0147] Missing value handling: For missing values ​​in user behavior data (such as records of a user not logging into the app), imputation is performed using the mean or mode. For missing values ​​in macroeconomic indicators (such as GDP data for a certain month not being released), interpolation is used or records for that time point are deleted.

[0148] Outlier Handling: For records where the policy amount significantly exceeds the industry average (e.g., the policy amount exceeds the 99th percentile of the industry), box plots or the 3σ rule are used to identify and remove outliers. For abrupt fluctuations in market trend data (e.g., a sudden increase in the premium growth rate in a certain month), a sliding window method is used for smoothing.

[0149] 3. Feature Engineering:

[0150] Feature extraction: Extract key features from user behavior data, such as user login frequency, surrender rate, and renewal rate; extract features from historical policy data, such as policy type, premium payment method, and loss ratio; extract features from macroeconomic indicators, such as market interest rate, GDP growth rate, and inflation rate; and extract features from market trend data, such as industry premium growth rate and competitor pricing strategies.

[0151] Feature encoding: For categorical variables (such as policy type, user occupation), one-hot encoding or label encoding is used.

[0152] Feature Combination: Combine user age and income level to construct user risk stratification features (e.g., high-income and young users have higher risk preferences); combine market interest rates and policy type to construct reserve cost prediction features (e.g., long-term insurance has higher reserve requirements in a low-interest-rate environment); Feature Normalization: For numerical features (e.g., premium amount, market interest rate), use Min-Max normalization or Z-Score standardization.

[0153] 4. Model Training and Optimization:

[0154] (1) Constructing a training set: Construct a training set using historical data and adjust the loss function based on the difference between the predicted results and the actual results.

[0155] (2) Parameter optimization: Use random search or Bayesian optimization methods to adjust model parameters to improve prediction accuracy.

[0156] (3) Performance evaluation: Use hold-out or bootstrap methods to evaluate the model performance and ensure its stability and reliability on different datasets.

[0157] Loss functions for model training: MSE (for regression tasks such as cost prediction) or cross-entropy (for classification tasks such as insurance withdrawal prediction).

[0158] Model training parameter optimization: adjusting hyperparameters (such as tree depth in random forests); Bayesian optimization is more efficient than grid search.

[0159] Evaluation methods for model training: hold-out method – 70% training / 30% testing; bootstrap method – resampling to evaluate model stability.

[0160] 5. Data Input Model:

[0161] The preprocessed feature data is input into the trained machine learning models, including the policy cost optimization model, the surrender cost prediction model, and the reserve cost optimization model.

[0162] (1) Policy cost calculation: Based on customer behavior data and market trend data, calculate the cost structure of different policy types and optimize policy design and pricing strategies.

[0163] Policy cost optimization: Customer income level (≤30k) + average market price of similar products → design differentiated pricing (e.g., a 5% reduction in monthly premium).

[0164] (2) Calculation of surrender costs: Based on the predicted surrender rate and surrender costs, optimize the policy renewal strategy and customer service strategy.

[0165] Policy surrender strategy optimization: Policies with a predicted surrender rate > 20% → automatically trigger renewal benefits (such as free supplementary insurance).

[0166] (3) Calculation of reserve cost: Based on macroeconomic indicators and market interest rate data, calculate the reserve demand and cost, and optimize the reserve management strategy.

[0167] Reserve optimization: The 10-year Treasury bond yield decreased by 50 basis points → the reserve provision ratio for long-term insurance was increased by 3%, ultimately resulting in the optimal policy design, pricing strategy and reserve management plan.

[0168] 6. Dynamic adjustment and optimization:

[0169] We monitor market changes and customer demand shifts in real time, regularly update model parameters and prediction results, and monitor data such as market interest rates and competitor pricing (via API for real-time access to the central bank / industry association databases). Simultaneously, we incrementally train the model with new data each month, automatically deploy updated parameters (e.g., using an AWS SageMaker pipeline), and set threshold alerts (e.g., triggering model retraining when the week-on-week increase in insurance surrender rate exceeds 10%).

[0170] Specifically, to achieve dynamic optimization of liability costs, a three-layer linkage mechanism needs to be built: a data perception layer, a model iteration layer, and a decision execution layer.

[0171] 1. At the data perception layer, the system connects in real time to the central bank's open market operation interest rate interface (such as SHIBOR / DR007) and the banking, insurance and credit industry database (which updates the competitor pricing table daily) via API. It also connects to the enterprise's internal business systems (such as the core underwriting platform and the surrender work order database). The system uses the Kafka stream processing engine to achieve second-level data synchronization. Abnormal data is processed in real time by Flink and then stored in the ClickHouse cluster.

[0172] 2. The model iteration layer adopts a hybrid training strategy: Online learning is implemented for high-frequency market data (such as interest rate fluctuations), and parameter fine-tuning is achieved through PyTorch's elastic weight solidification technology; full batch training is triggered on the 1st of each month, and AWS SageMaker Pipelines are used for automated scheduling. After pulling incremental data from S3, the XGBoost / Transformer model is retrained, and the differences between the old and new models in terms of surrender rate prediction (MAPE≤5%) and reserve coverage ratio (error±1.5%) are compared through A / B testing. After the targets are met, the model is automatically deployed to the production environment through the MLflow model repository.

[0173] 3. Intelligent threshold rules are set at the decision-making and execution layer: When the week-on-week increase in the surrender rate exceeds 10%, the system automatically freezes the original model and rolls back to the previous stable version. Simultaneously, a cross-departmental collaborative process is triggered—the actuarial department adjusts liability reserve assumptions, and the operations department pushes targeted renewal incentives (such as "one year's premium waived for three consecutive years of renewal"). All strategy changes are recorded through the Jira ticket system and synchronized to the regulatory reporting engine to ensure compliance with the "Solvency II" transparent regulatory requirements. After this architecture was implemented at a leading life insurance company, the model iteration cycle was reduced from 45 days to 6 hours, and the surrender warning response speed increased by 17 times. Ultimately, it achieved precise optimization of insurance product design and pricing strategies, thereby reducing liability-side costs.

[0174] For example, an insurance company designed a whole life insurance product for a 35-year-old high-income user (annual income of approximately 1 million yuan). The user's insurance records show that they tend to choose products with high coverage and low premiums, and have a high risk appetite (preferring to take on higher risks to obtain higher returns).

[0175] Specific applications are as follows:

[0176] 1. Data Acquisition:

[0177] Historical policy data: This user previously purchased an annuity insurance policy with a coverage of 500,000 yuan, a payment period of 10 years, and an annual premium of 50,000 yuan; User behavior data: This user logged into the App 20 times in the past 12 months, participated in 3 online insurance consultations, and did not cancel any policies; Risk preference data: The user selected "willing to take on higher risks to obtain higher returns" in the risk tolerance questionnaire; Macroeconomic indicators: The current 10-year treasury bond yield is 3.5%, the GDP growth rate is 5.2%, and the inflation rate is 2.1%; Market trend data: The average premium for similar whole life insurance products is 60,000 yuan / year, and the industry surrender rate is 5%.

[0178] 2. Data preprocessing:

[0179] Missing value handling: There are no missing values ​​in user behavior data and macroeconomic indicators; Outlier handling: The user's policy amount is 500,000 yuan, which is lower than the 99th percentile (1 million yuan) of the industry, and there are no outliers.

[0180] Feature extraction: User behavior characteristics: login frequency 20 times / year, surrender rate 0%; Risk preference characteristics: high risk preference; Macroeconomic characteristics: market interest rate 3.5%, GDP growth rate 5.2%; Market trend characteristics: competitors' average premium is 60,000 yuan / year, industry surrender rate 5%.

[0181] 3. Feature Engineering:

[0182] Feature encoding:

[0183] Risk preference: High risk preference → 1, medium risk preference → 0, low risk preference → -1.

[0184] Feature combination: User age (35 years old) + income level (1 million yuan) → high risk tolerance; market interest rate (3.5%) + policy type (whole life insurance) → high reserve requirement.

[0185] 4. Model Input and Output:

[0186] Policy cost optimization model: Based on user risk preferences and market trends, it is recommended to design a whole life insurance policy with high coverage and low premiums, with an annual premium of 55,000 yuan; Surrender cost prediction model: The predicted surrender rate is 3%, and it is recommended to offer renewal discounts (such as free supplementary insurance); Reserve cost optimization model: Based on market interest rates and policy type, it is recommended to increase the reserve provision ratio to 5%.

[0187] 5. Optimization results:

[0188] Final policy plan: Annual premium of RMB 55,000, coverage includes death, total disability, and critical illness. Renewal discount: 50% premium reduction in the 3rd year. Reserve provision ratio is 5%. Total liability cost: Policy cost + Surrender cost + Reserve cost = RMB 55,000 + RMB 1,650 + RMB 2,750 = RMB 58,425 per year.

[0189] Ultimately, through model optimization in this embodiment, the user's total policy cost was reduced by approximately 10%, while maintaining a high level of customer satisfaction. By dynamically adjusting pricing strategies and reserve management, the insurance company significantly improved cost management efficiency and market competitiveness on the liability side.

[0190] Figure 8 This is a schematic diagram of the business processing device provided in the embodiments of this application. Please refer to it. Figure 8 This application embodiment also provides a business processing apparatus 800, which can implement the above-described business processing method. The business processing apparatus 800 includes:

[0191] The response module 801 is used to respond to the target user's request to process the target business, and to obtain the market economic data of the target business and the characteristic data of the target user; the characteristic data includes user behavior data, historical business data and risk preference data;

[0192] The generation module 802 is used to generate at least two feature combinations based on the data in the market economy data and feature data;

[0193] The processing module 803 is used to input at least two feature combinations into the business processing model, and through each sub-model of the business processing model, determine the business indicators corresponding to each sub-model based on the feature combinations corresponding to each sub-model from the at least two feature combinations, and generate a business processing plan corresponding to the target business and the target user based on the business indicators corresponding to each sub-model.

[0194] In some embodiments, the processing module 803 includes:

[0195] The first input module is used to input the first feature combination into the policy cost optimization model to obtain the policy cost index output by the policy cost optimization model; the first feature combination is generated based on market economic data, user behavior data of target users, and risk preference data;

[0196] The second input module is used to input the second feature combination into the surrender cost prediction model to obtain the surrender cost index output by the surrender cost prediction model; the second feature combination is generated based on the target user's user behavior data, historical business data and risk preference data;

[0197] The third input module is used to input the third feature combination into the reserve cost optimization model to obtain the reserve cost index output by the reserve cost optimization model; the third feature combination is generated based on market economic data and user behavior data of target users.

[0198] In some embodiments, the processing module 803 further includes any one of the following:

[0199] The data acquisition module is used to acquire historical insurance policy datasets;

[0200] The first building module is used to build an initial policy cost optimization model based on the random forest model;

[0201] The second building module is used to construct an initial surrender cost prediction model using a gradient boosting tree with a temporal attention mechanism.

[0202] The third building module is used to build an initial reserve cost optimization model based on a preset deep learning model.

[0203] The model optimization module is used to train the initial policy cost optimization model, the initial surrender cost prediction model, and the initial reserve cost optimization model based on historical insurance policy datasets, respectively, to obtain the policy cost optimization model, the surrender cost prediction model, and the reserve cost optimization model.

[0204] In some embodiments, the processing module 803 further includes any one of the following:

[0205] The calculation module is used to calculate the comprehensive business indicators of the target business based on the policy cost indicator, surrender cost indicator, and reserve cost indicator.

[0206] The first generation submodule is used to generate a renewal discount plan for the target business when the surrender cost indicator is less than or equal to the preset surrender cost indicator threshold.

[0207] The second generation submodule is used to generate a business processing plan for the target business based on the renewal discount scheme and the comprehensive business indicators of the target business.

[0208] In some embodiments, the response module 801 includes:

[0209] The first acquisition module is used to acquire initial macroeconomic data, initial market trend data, initial user behavior data, initial historical business data, and initial risk preference data for the target business.

[0210] The first processing submodule is used to perform anomaly processing on the initial macroeconomic data, initial user behavior data, and initial risk preference data to obtain macroeconomic data, user behavior data, and risk preference data; the anomaly processing includes at least missing value handling and outlier handling.

[0211] The second processing submodule is used to smooth the initial market trend data using the sliding window method to obtain the market trend data.

[0212] The third processing submodule is used to process the initial historical business data using box plots or the Raida criterion to obtain historical business data.

[0213] In some embodiments, the generation module 802 includes:

[0214] The extraction module is used to extract data features from macroeconomic data, market trend data, user behavior data, historical business data, and risk preference data, respectively, to obtain macroeconomic features, market trend features, user behavior features, historical business features, and risk preference features.

[0215] The first combination module is used to combine user behavior characteristics, market trend characteristics, and risk preference characteristics to obtain the first feature combination.

[0216] The second combination module is used to combine user behavior characteristics, historical business characteristics, and risk preference characteristics to obtain a second feature combination.

[0217] The third combination module is used to combine user behavior characteristics, macroeconomic characteristics, and market trend characteristics to obtain the third feature combination.

[0218] Among them, at least two feature combinations include at least a first feature combination, a second feature combination, and a third feature combination.

[0219] In some embodiments, the processing module 803 further includes any one of the following:

[0220] The second acquisition module is used to acquire market interest rate fluctuation signals of the target business, competitor pricing strategy change events, and newly added historical business datasets, provided that the target conditions are met.

[0221] The parameter update module is used to update the parameters of the policy cost optimization model, surrender cost prediction model and reserve cost optimization model based on market interest rate fluctuation signals, competitor pricing strategy change events and new historical business datasets.

[0222] The target conditions include: the comprehensive business indicator generated based on each business indicator is greater than or equal to the preset business indicator threshold.

[0223] The specific implementation of the business processing device 800 is basically the same as the specific implementation of the business processing method described above, and will not be repeated here.

[0224] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described business processing method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0225] Please see Figure 9 , Figure 9 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes:

[0226] The processor 901 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0227] The memory 902 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called and executed by the processor 901 using the business processing methods of the embodiments of this application.

[0228] The input / output interface 903 is used to implement information input and output;

[0229] The communication interface 904 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0230] Bus 905 transmits information between various components of the device (e.g., processor 901, memory 902, input / output interface 903, and communication interface 904);

[0231] The processor 901, memory 902, input / output interface 903, and communication interface 904 are connected to each other within the device via bus 905.

[0232] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described business processing method.

[0233] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0234] The business processing method, business processing apparatus, electronic device, and medium provided in this application embodiment respond to a target user's request for processing a target business by acquiring market economic data of the target business and characteristic data of the target user; generating at least two feature combinations based on each data in the market economic data and characteristic data; inputting the feature combinations into a business processing model; determining the business indicators corresponding to each sub-model based on the feature combinations corresponding to each sub-model in the feature combinations; and generating a business processing scheme corresponding to the target business and the target user based on the business indicators corresponding to each sub-model. This achieves dynamic optimization and accurate decision-making of the business processing scheme by dynamically integrating market economic data and multi-dimensional user characteristics, constructing multi-dimensional feature combinations, and utilizing sub-models for collaborative processing.

[0235] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0236] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0237] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0238] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0239] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0240] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0241] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above 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 coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0242] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0243] 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.

[0244] If the integrated 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, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0245] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A business processing method, characterized in that, Applied to an electronic device, the electronic device including a business processing model, the business processing model including at least one sub-model, the method includes: In response to a target user's request to process a target service, the system acquires market economic data of the target service and characteristic data of the target user; the characteristic data includes user behavior data, historical business data, and risk preference data. Based on the market economy data and the data in the feature data, generate at least two feature combinations; The at least two feature combinations are input into the business processing model. Based on the feature combinations corresponding to each sub-model in the at least two feature combinations, the business indicators corresponding to each sub-model are determined by each sub-model. Based on the business indicators corresponding to each sub-model, a business processing plan corresponding to the target business and the target user is generated.

2. The business processing method according to claim 1, characterized in that, When the target business is insurance policy business, the business processing model includes a policy cost optimization model, a surrender cost prediction model, and a reserve cost optimization model. The process involves inputting the at least two feature combinations into the business processing model, and then determining the corresponding business indicators for each sub-model based on the feature combinations corresponding to the at least two feature combinations. This includes: The first feature combination is input into the policy cost optimization model to obtain the policy cost index output by the policy cost optimization model; the first feature combination is generated based on the market economic data, the target user's user behavior data, and risk preference data. The second feature combination is input into the surrender cost prediction model to obtain the surrender cost index output by the surrender cost prediction model; the second feature combination is generated based on the target user's user behavior data, historical business data, and risk preference data. The third feature combination is input into the reserve cost optimization model to obtain the reserve cost index output by the reserve cost optimization model; the third feature combination is generated based on market economic data and user behavior data of the target user.

3. The business processing method according to claim 2, characterized in that, Before inputting the at least two feature combinations into the business processing model, and determining the business indicators corresponding to each sub-model based on the feature combinations corresponding to each sub-model from the at least two feature combinations, the method further includes: Obtain historical insurance policy dataset; Based on the random forest model, construct an initial policy cost optimization model; A gradient boosting tree with temporal attention mechanism is used to construct an initial surrender cost prediction model; Based on the pre-defined deep learning model, construct an initial reserve cost optimization model; Based on the historical insurance policy dataset, the initial policy cost optimization model, the initial surrender cost prediction model, and the initial reserve cost optimization model are trained to obtain the policy cost optimization model, the surrender cost prediction model, and the reserve cost optimization model.

4. The business processing method according to claim 3, characterized in that, The step of generating a business processing plan corresponding to the target business and the target user based on the business metrics corresponding to each sub-model includes: The comprehensive business indicators for the target business are calculated based on the policy cost indicator, the surrender cost indicator, and the reserve cost indicator. If the surrender cost index is less than or equal to a preset surrender cost index threshold, a renewal discount scheme for the target business is generated. Based on the renewal discount scheme, a business processing scheme for the target business is generated according to the comprehensive business indicators of the target business.

5. The business processing method according to claim 2, characterized in that, Before acquiring the market economic data of the target business and the characteristic data of the target user, the method further includes: Acquire the initial macroeconomic data, initial market trend data, initial user behavior data, initial historical business data, and initial risk preference data for the target business; The initial macroeconomic data, the initial user behavior data, and the initial risk preference data are subjected to anomaly processing to obtain the macroeconomic data, user behavior data, and risk preference data; the anomaly processing includes at least missing value processing and outlier processing; The initial market trend data is smoothed using a sliding window method to obtain the market trend data. The initial historical business data is processed using a box plot or the Raida criterion to obtain the historical business data.

6. The business processing method according to claim 2, characterized in that, The market economic data includes macroeconomic data and market trend data. The step of generating at least two feature combinations based on the market economic data and the feature data includes: Data features are extracted from the macroeconomic data, market trend data, user behavior data, historical business data, and risk preference data to obtain macroeconomic features, market trend features, user behavior features, historical business features, and risk preference features, respectively. The user behavior characteristics, the market trend characteristics, and the risk preference characteristics are combined to obtain the first feature combination; The user behavior characteristics, historical business characteristics, and risk preference characteristics are combined to obtain the second feature combination; By combining the user behavior characteristics, the macroeconomic characteristics, and the market trend characteristics, a third characteristic combination is obtained; The at least two feature combinations include at least the first feature combination, the second feature combination, and the third feature combination.

7. The business processing method according to claim 1, characterized in that, After generating the business processing plan corresponding to the target business and the target user based on the business metrics corresponding to each sub-model, the method includes: Under the condition that the target conditions are met, obtain the market interest rate fluctuation signal of the target business, the competitor pricing strategy change event, and the newly added historical business dataset; Based on the market interest rate fluctuation signal, the competitor pricing strategy change event, and the newly added historical business dataset, the parameters of the policy cost optimization model, the surrender cost prediction model, and the reserve cost optimization model are updated. The target conditions include: the comprehensive business indicator generated based on each of the business indicators is greater than or equal to the preset business indicator threshold.

8. A business processing device, characterized in that, The device includes: The response module is used to respond to a target user's request to process a target service, and to obtain market economic data of the target service and characteristic data of the target user; the characteristic data includes user behavior data, historical business data and risk preference data. The generation module is used to generate at least two feature combinations based on the market economy data and each data in the feature data; The processing module is used to input the at least two feature combinations into the business processing model, determine the business indicators corresponding to each sub-model based on the feature combinations corresponding to each sub-model among the at least two feature combinations, and generate the business processing scheme corresponding to the target business and the target user based on the business indicators corresponding to each sub-model.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the business processing method according to 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 the processor, it implements the business processing method according to any one of claims 1 to 7.

Images