A service early warning method, device, equipment and medium

Abstract

The application discloses a business early warning method, device, equipment and medium, which can be applied to the financial field, and comprises the following steps: determining a business score of a customer in a business scene based on identity information and historical behavior information of the customer; determining common business characteristics and different business characteristics of a business institution based on institution business data of a plurality of different business institutions corresponding to the business scene; determining a business early warning threshold of the customer in the business scene based on the common business characteristics and the different business characteristics and a business institution to which the customer belongs; when the business score exceeds the business early warning threshold, determining a target business early warning mode of the customer based on a business level corresponding to the business scene and the historical behavior information, and sending target early warning information through the target early warning mode. The method jointly analyzes data of different business institutions, extracts common and different characteristics, dynamically determines a business early warning threshold, and improves the accuracy and adaptability of the early warning threshold.

Description

Technical Field

[0001] This invention relates to the field of business early warning, and in particular to a business early warning method, apparatus, equipment and medium. Background Technology

[0002] Currently, across various industries, early warning systems typically rely on fixed thresholds and customer tag matching for business scenarios. For example, in the financial sector, auto insurance often triggers an alert when a customer's accident count exceeds a certain threshold. However, such methods overlook the dynamic impact of commonalities and differences among different business entities, making it difficult to adapt to the complexity and diversity of customer behavior in different regions, thus limiting the accuracy of early warnings. Therefore, there is an urgent need for a dynamic business early warning method that can integrate the commonalities and differences among multiple institutions. Summary of the Invention

[0003] This invention provides a business early warning method, apparatus, device, and medium to solve the problem of low accuracy of business early warning caused by fixed thresholds and static tags in related technologies.

[0004] Firstly, this disclosure provides a business early warning method, including: Based on customer identity information and historical behavior information, determine the customer's business score in the business scenario; Based on the business data of multiple different business units corresponding to business scenarios, the common and different business characteristics of the business units are determined. Based on common and different business characteristics, as well as the business organization to which the customer belongs, determine the business warning threshold for the customer in the business scenario. When the business score exceeds the business warning threshold, the target business warning method for the customer is determined based on the business level and historical behavior information of the corresponding business scenario, and the target warning information is sent through the target warning method.

[0005] Secondly, this disclosure provides a business early warning device, comprising: The rating determination module is used to determine the customer's business rating in a business scenario based on the customer's identity information and historical behavior information. The feature determination module is used to determine the common and differential business features of business units based on the organizational business data of multiple different business units corresponding to business scenarios. The threshold determination module is used to determine the business warning threshold for a customer in a business scenario based on common business characteristics, different business characteristics, and the business organization to which the customer belongs. The early warning execution module is used to determine the target business early warning method for the customer based on the business level and historical behavior information of the corresponding business scenario when the business score exceeds the business early warning threshold, and then send the target early warning information through the target early warning method.

[0006] Thirdly, this disclosure provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned business early warning method. Fourthly, this disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned business early warning method.

[0007] The aforementioned business early warning method, device, equipment, and medium implement a solution that determines a customer's business score in a business scenario based on the customer's identity information and historical behavior information; determines the common and differential business characteristics of the business organizations based on organizational business data corresponding to the business scenario; determines the customer's business early warning threshold in the business scenario based on the common and differential business characteristics and the business organization to which the customer belongs; when the business score exceeds the business early warning threshold, determines the customer's target business early warning method based on the business level of the corresponding business scenario and the historical behavior information, and sends the target early warning information through the target early warning method. This method, by jointly analyzing the organizational data of different business organizations and extracting common and differential characteristics, achieves dynamic determination of the business early warning threshold, improving the accuracy and adaptability of the early warning threshold setting, effectively reducing the false alarm rate and false negative rate. Simultaneously, by combining business registration and historical behavior information, it dynamically selects an appropriate target early warning method to choose the appropriate early warning channel for early warning, thereby improving early warning efficiency. Attached Figure Description

[0008] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a flowchart of a business early warning method according to an embodiment of the present invention; Figure 2 This is another flowchart of a business early warning method in one embodiment of the present invention; Figure 3 This is another flowchart of a business early warning method in one embodiment of the present invention; Figure 4 This is another flowchart of a business early warning method in one embodiment of the present invention; Figure 5 This is another flowchart of a business early warning method in one embodiment of the present invention; Figure 6 This is a schematic block diagram of a business early warning device in one embodiment of the present invention; Figure 7 This is a schematic block diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0010] In one embodiment, such as Figure 1 As shown, a business early warning method is provided, including the following steps: S101, Based on the customer's identity information and historical behavior information, determine the customer's business score in the business scenario; S102, Based on the business data of multiple different business units corresponding to business scenarios, determine the common business characteristics and different business characteristics of the business units. S103, based on common business characteristics and different business characteristics, as well as the business organization to which the customer belongs, determine the business warning threshold for the customer in the business scenario; S104. When the business score exceeds the business warning threshold, the target business warning method for the customer is determined based on the business level and historical behavior information of the corresponding business scenario, and the target warning information is sent through the target warning method.

[0011] As an example, in step S101, the customer's identity information includes static information such as the customer's age, occupation, channel (online channel, agent channel, etc.), and region, while the historical behavior information includes dynamic data such as the customer's browsing history on the platform, search keywords, access frequency and duration, and insurance intention behavior.

[0012] Furthermore, different model algorithms can be used to calculate customer identity information and historical behavior information based on different business scenarios to generate corresponding business scores.

[0013] For example, in the renewal scenario in the financial sector, the XGBoost model is used to process customer identity information and historical behavior information to obtain the customer's renewal probability, and this renewal probability is quantified to obtain the corresponding business score.

[0014] For example, in the anti-fraud scenario of insurance, deep correlation analysis is performed on customers' historical behavior information (such as historical insurance behavior) and identity information (such as associated customer information) based on graph neural networks to identify the connection path between them and known fraud networks, output the corresponding fraud risk probability, and quantify it into a business score.

[0015] As an example, in step S102, by performing cluster analysis on the business data of each business unit, the common and different characteristics of each business unit in terms of customer structure, business preferences and risk distribution are identified. For example, the customer groups of some institutions generally have a high risk preference, while other institutions have fewer high-risk businesses.

[0016] For example, a federated learning framework can be used to jointly learn from the local data uploaded by various business units. Without disclosing the original data of each business unit, a global common feature model can be constructed. This global model can then be used to extract common business features from each business unit. Simultaneously, by comparing the residual features of each business unit based on the output of the global model, the unique business features of each business unit can be identified. The aforementioned business units may be, for example, insurance institutions, etc., but this disclosure does not limit this to such cases.

[0017] As an example, in step S103, by weighting and fusing common business characteristics and different business characteristics, and setting dynamic adjustment factors in conjunction with business scenarios, personalized early warning thresholds for different business organizations and different business scenarios are generated, so as to realize the dynamic determination of business early warning thresholds and avoid the problems of misjudgment or missed judgment caused by fixed thresholds.

[0018] As an example, in step S104, when a customer's business score exceeds the business warning threshold of the business scenario corresponding to their business unit, a risk warning mechanism is triggered. Based on the preset risk level mapping relationship of the business scenario, the corresponding warning level is determined, and the warning method is dynamically selected in combination with the customer's historical behavior information.

[0019] For example, for high-risk warnings, historical behavioral information can be used to identify channels with faster customer response times, such as instant messaging tools or telephones, for real-time alerts; for medium- and low-risk warnings, historical behavioral information can be used to select channels that customers frequently use and that have stable responses, such as SMS or emails, for scheduled alerts, to ensure that the warning information is effectively delivered.

[0020] In summary, this disclosure proposes a business early warning method, comprising: determining a customer's business score in a business scenario based on the customer's identity information and historical behavior information; determining common and differential business characteristics of the business institutions based on institutional business data corresponding to the business scenario; determining a business early warning threshold for the customer in the business scenario based on the common and differential business characteristics and the business institution to which the customer belongs; and determining a target business early warning method for the customer based on the business level of the corresponding business scenario and the historical behavior information when the business score exceeds the business early warning threshold, and sending a target early warning message through the target early warning method. This method, by jointly analyzing institutional data from different business institutions and extracting common and differential characteristics, dynamically determines the business early warning threshold, improving the accuracy and adaptability of the early warning threshold setting, effectively reducing the false alarm rate and false negative rate. Furthermore, by combining business registration and historical behavior information, it dynamically selects an appropriate target early warning method to choose the appropriate early warning channel for early warning, thereby improving early warning efficiency.

[0021] In one embodiment, such as Figure 2 As shown, step S102, which involves determining the common and differential business characteristics of business units based on the organizational business data of multiple different business units corresponding to business scenarios, includes: S201, Based on institutional business data, determine the local institutional characteristics corresponding to the business scenario; S202, based on a joint comparison platform, performs feature alignment on local institutional features to obtain cross-institutional shared features: S203, based on a global sharing model, performs correlation analysis on cross-organizational shared features to obtain common business characteristics: S204. Compare local organizational characteristics and cross-organizational shared characteristics to determine differentiated business characteristics.

[0022] As an example, in step S201, the institutional business data refers to the data of all business scenarios in the business institution. By decomposing and extracting features from the institutional business data, local feature vectors related to the current business scenario are obtained.

[0023] For example, in the scenario of auto insurance renewal, the local institution's feature vector can be constructed by extracting dimensions such as the customer's insurance period, quote frequency, and claims record with the local insurance company.

[0024] As an example, in step S202, the local institutional characteristics of each insurance company are standardized and aligned through a joint comparison platform to eliminate data heterogeneity, identify cross-institutional shared feature patterns, and form cross-institutional shared features.

[0025] In other words, shared institutional features are the set of comparable and aggregateable features exhibited by various institutions under the same business scenario. This means that by aligning the local institutional features of each institution and mapping them to the same feature space, a set of feature vectors is formed, ensuring the consistency and comparability of feature expressions among different institutions.

[0026] For example, taking the scenario of car insurance renewal as an example, common features such as pricing strategies and renewal time preferences of insurance companies in different regions can be standardized and incorporated into a cross-institutional shared feature vector.

[0027] As an example, in step S203, the correlation analysis of cross-organization shared features can be performed based on the global sharing model to explore the potential dependencies and collaboration patterns among the features, and then extract stable and predictive common business features.

[0028] For example, by analyzing the cross-institutional shared features of multiple insurance institutions, a strong positive correlation was found between the renewal time preference feature and the quotation frequency feature, and this pattern remained stable in different regions. Thus, the renewal time preference feature and the quotation frequency feature can be identified as common business features in the auto insurance renewal scenario.

[0029] As an example, in step S204, by comparing local organizational features with cross-organizational shared features, features that exist only in specific business organizations are identified and determined as differentiated business features.

[0030] For example, by comparing the local institutional characteristics and cross-institutional shared characteristics of multiple financial institutions, it was found that the high-risk financial product preference characteristics of a certain local financial institution's customers were not reflected in the common business characteristics. Therefore, the high-risk financial product preference characteristics can be identified as differentiated business characteristics, reflecting the uniqueness of the institution in the behavior of its customer group, which can lay the foundation for determining the subsequent personalized business early warning threshold.

[0031] In one embodiment, such as Figure 3 As shown, step S103, which involves determining the customer's business warning threshold in the business scenario based on common and differentiated business characteristics and the business organization to which the customer belongs, includes: S301, based on the business organization to which the customer belongs, integrates common and different business characteristics corresponding to the business scenarios to obtain integrated business characteristics; S302, based on the characteristics of converged services, determines the service early warning threshold.

[0032] As an example, in step S301, a fusion business feature vector is constructed based on common business characteristics and the different business characteristics of the customer's affiliated institution to ensure the coordinated expression of common risk patterns and institution-specific behavioral patterns.

[0033] For example, in the scenario of auto insurance renewal, common business characteristics such as renewal time preference and quotation frequency are weighted and integrated with a certain institution's unique tendency to insure high-risk vehicle models to form a personalized characteristic representation for the institution's customers, providing a basis for the accurate setting of subsequent early warning thresholds.

[0034] When fusing common business features and different business features, an adaptive weight allocation mechanism can be adopted to dynamically adjust their contribution based on the predictive power of common business features and different business features in order to improve the discrimination ability of the fused features. However, it is not limited to this and fixed weights can also be set in combination with expert experience.

[0035] As an example, in step 302, the business warning threshold can be automatically learned and determined based on the fused business feature vector, using a machine learning model or threshold optimization algorithm, combined with the warning feedback results in the organization's business data.

[0036] For example, by using logistic regression or XGBoost models to model the integrated business feature vector of customer renewal business, and combining it with actual default or churn tag data from institutional business data, the boundary of the risk warning threshold can be determined to identify the corresponding risk warning threshold. This method not only takes into account common patterns across institutions and local institutional characteristics, but also adapts to dynamic changes in different business scenarios, enhancing the timeliness and personalization of warning identification.

[0037] In one embodiment, such as Figure 4 As shown, step S104, which is to determine the customer's target business alert method based on the business level and historical behavior information of the corresponding business scenario: S401, based on business level, determines the business value of a business scenario; S402, based on historical behavior information, determine the customer's response time among multiple candidate business alert methods; S403 determines the target response method from multiple candidate warning methods based on business value and response time.

[0038] As an example, in step S401, business levels can be pre-classified according to the business benefits and risk impact of different business scenarios (e.g., divided into high-value scenarios, medium-value scenarios, and low-value scenarios, but not limited to this); and based on the business level of the business scenario, the business value of the corresponding business scenario can be determined according to the pre-set mapping relationship between business level and business value.

[0039] As an example, in step S402, by analyzing the customer's historical behavior information, the customer's response time to different candidate business warning methods (such as the viewing time of SMS, APP push or telephone reminders) is obtained and the average is calculated. At the same time, based on the historical behavior information, the predicted response time of the customer to different warning methods is predicted. Then, the average calculation and the predicted response time are weighted and fused, and the fused result is determined as the target response method.

[0040] For example, by using the Transformer model to model historical behavioral information, we can capture the sequence patterns of customer response preferences to various warning methods at different time periods, and predict the predicted response time of customers to different warning methods.

[0041] As an example, in step S403, a weighted scoring model is used to comprehensively evaluate the business value and the customer's response time to each warning method, and the warning method with the best response time in high-value scenarios is selected first; for cases with the same response time, the customer's historical preferences and reach costs are further considered, and the strategy is dynamically adjusted to achieve accurate reach and optimized resource allocation.

[0042] For example, when issuing business policy warnings to customers, channels such as WeChat and SMS can be selected as the target response methods; when issuing anti-fraud warnings to customers, pop-up reminders in the user's business APP or telephone reminders by business personnel can be selected as the target response methods.

[0043] In one embodiment, such as Figure 5 As shown, step S104, which involves sending target warning information, includes: S501, based on historical behavior information, obtains historical warning dialogue information of customers in business scenarios; S502 generates candidate warning information based on the target response method and business scenario through a natural language model; S503 modifies candidate warning information based on historical warning dialogue information, generates target warning information, and sends the target warning information to the customer.

[0044] As an example, in step S501, historical warning dialogue information of customers in similar business scenarios in the past can be extracted from historical behavior information. The historical warning dialogue information may include information such as the customer's feedback on the warning information, response time and number of interaction rounds.

[0045] As an example, in step S502, a pre-trained natural language generation model is used to generate candidate warning information text that conforms to the customer's understanding habits by combining the semantic features of the current business scenario and the channel characteristics of the target response method.

[0046] For example, in the scenario of car insurance renewal reminders, if the target response method is SMS, a concise and clear short text is generated, including the expiration date, discount information, and operation link, such as "Your car insurance will expire in 3 days. Enjoy a 20% discount on renewal. Click the link to process now." In the scenario of anti-fraud warnings, if the target response method is an app pop-up, a highly suggestive text is generated, highlighting risk warnings and emergency operation guidance, such as "Abnormal account login behavior detected! Please verify and change your password immediately to protect your funds."

[0047] As an example, in step S503, the candidate warning information is personalized and adjusted by combining the feedback tendency and text style preference of the customer's historical warning dialogues to ensure that the semantics are clear and in line with the customer's cognitive habits. Finally, the target warning information is generated and sent to improve the customer's willingness to respond and the interactive experience.

[0048] For example, if historical alert dialogue information shows that customers respond more readily to formal reminders, then rigorous wording should be prioritized when generating target alert information; if customers prefer concise expressions, then the content should be streamlined to highlight key information.

[0049] It should be understood that when sending target warning information, it may be sent to the customer or to relevant personnel of the customer's associated agent or business unit to coordinate risk handling or service response, and this disclosure does not limit it in this way.

[0050] In an optional embodiment, before step S101, i.e., determining the customer's business score in the business scenario based on the customer's identity information and historical behavior information, the method further includes: obtaining the customer's business data within a preset time period; determining whether the customer meets the warning triggering conditions based on the business data; and obtaining the identity information and historical behavior information when the customer meets the warning triggering conditions.

[0051] The preset time period can be the past 30 days or dynamically adjusted according to the business scenario. The business data includes key indicators such as transaction frequency, amount changes, and login behavior, so as to determine whether to issue an early warning to the customer through multi-dimensional data analysis.

[0052] In one embodiment, a preset early warning rule engine performs real-time calculations and comparisons on business data to determine whether a customer meets the early warning triggering conditions. If the triggering conditions are met, the early warning process is initiated immediately to ensure timely response. If the early warning triggering conditions are not met, the system continuously monitors the customer's behavioral data to avoid missing high-risk behaviors or causing frequent information prompts to the customer.

[0053] For example, if the warning trigger condition is "the customer visits the car insurance page for two consecutive days but does not purchase insurance, and has searched for the keyword 'new energy vehicle insurance'", then the rule engine monitors the user behavior log in real time. When it detects that the access frequency and keyword search behavior meet the conditions at the same time, it immediately triggers the warning process, obtains the customer's identity information and historical behavior information, and determines whether to issue a business warning to the relevant business personnel to prompt them to follow up with potential customers who intend to purchase insurance.

[0054] In an optional embodiment, after step S104, i.e., after sending the target warning information, the method further includes: obtaining the response operation corresponding to the target warning information; determining the business conversion parameters of the target warning information based on the response operation and through a causal model; and updating the warning triggering conditions and / or business warning thresholds based on the business conversion parameters.

[0055] Specifically, by statistically analyzing the response actions of customers or corresponding business personnel (such as clicking to respond, following up, or ignoring), the actual effect of the warning information is determined. Then, a causal inference model (such as Causal Forest) is used to analyze the degree of its impact on business conversion (such as renewal rate and insurance rate). The degree of impact is then quantified to obtain business conversion parameters. Based on these parameters, it is determined whether the warning triggering conditions and / or business warning thresholds need to be dynamically adjusted to avoid the risk of overloaded warnings or missed warnings.

[0056] Specifically, when business conversion parameters show that the sending of the target warning information has a significant positive impact, the business warning threshold can be tightened to improve sensitivity; if business conversion parameters indicate that the sending of the target warning information is ineffective or the response rate is consistently low, the range of triggering conditions can be expanded or the priority can be reduced to achieve adaptive iteration of the model. Ultimately, through a data feedback loop, the risk identification capability is continuously improved, maximizing business value while ensuring user experience.

[0057] In an optional embodiment, a visual front-end interface can also be configured on the execution device of the above-mentioned business early warning method. This interface allows business personnel to customize early warning triggering conditions and business early warning thresholds through drag-and-drop, and displays simulated curves of early warning coverage and reach frequency after the early warning triggering conditions and business early warning thresholds take effect in real time. Business personnel can dynamically optimize the early warning strategy configuration based on the simulated data on the visual interface, improving the scientific nature and flexibility of rule formulation. The system synchronously records the operation log of each configuration change and associates it with subsequent early warning effect data to form a traceable strategy iteration path.

[0058] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0059] In one embodiment, a service early warning device is provided, which corresponds one-to-one with the service early warning method in the above embodiments. For example... Figure 6 As shown, the business early warning device includes a scoring determination module 601, a feature determination module 602, a threshold determination module 603, and an early warning execution module 604. Detailed descriptions of each functional module are as follows: The scoring determination module 601 is used to determine the customer's business score in the business scenario based on the customer's identity information and historical behavior information. The feature determination module 602 is used to determine the common business features and differential business features of business units based on the organizational business data of multiple different business units corresponding to business scenarios. The threshold determination module 603 is used to determine the business warning threshold of a customer in a business scenario based on common business characteristics, different business characteristics, and the business organization to which the customer belongs. The early warning execution module 604 is used to determine the customer's target business early warning method based on the business level and historical behavior information of the corresponding business scenario when the business score exceeds the business early warning threshold, and to send the target early warning information through the target early warning method.

[0060] In one embodiment, the feature determination module 602 is further configured to determine local institutional features corresponding to the business scenario based on institutional business data. Based on a joint comparison platform, feature alignment is performed on local institutional features to obtain cross-institutional shared features: Based on a global sharing model, correlation analysis is performed on cross-organizational shared features to obtain common business characteristics: By comparing local institutional characteristics with cross-institutional shared characteristics, differentiated business characteristics can be identified.

[0061] In one embodiment, the threshold determination module 603 is further configured to obtain integrated business features by integrating common business features and differential business features corresponding to the business scenario based on the business organization to which the customer belongs; Based on the characteristics of integrated services, business early warning thresholds are determined.

[0062] In one embodiment, the early warning execution module 604 is further configured to determine the business value of a business scenario based on the business level; Based on historical behavioral information, determine the customer's response time among multiple candidate business alert methods; Based on business value and response time, the target response method is determined from multiple candidate early warning methods.

[0063] In one embodiment, the early warning execution module 604 is further configured to obtain historical early warning dialogue information of the customer in the business scenario based on historical behavior information; Based on the target response method and business scenario, candidate early warning information is generated through natural language models; Based on historical early warning dialogue information, candidate early warning information is modified to generate target early warning information, which is then sent to the customer.

[0064] In one embodiment, the scoring determination module 601 is further configured to acquire customer business data within a preset time period; Based on business data, determine whether the customer meets the warning trigger conditions; When a customer meets the alert trigger conditions, their identity information and historical behavior information are obtained.

[0065] In one embodiment, the early warning execution module 604 is further configured to acquire the response operation corresponding to the target early warning information; Based on response operations, the business conversion parameters of the target early warning information are determined through a causal model; Based on business conversion parameters, update the early warning trigger conditions and / or business early warning thresholds.

[0066] This invention provides a business early warning device, comprising: a scoring determination module for determining a customer's business score in a business scenario based on the customer's identity information and historical behavior information; a feature determination module for determining common and differential business features of business organizations based on organizational business data from multiple different business organizations corresponding to the business scenario; a threshold determination module for determining a business early warning threshold for the customer in the business scenario based on the common and differential business features and the business organization to which the customer belongs; and an early warning execution module for determining the customer's target business early warning method based on the business level and historical behavior information of the corresponding business scenario when the business score exceeds the business early warning threshold, and sending target early warning information through the target early warning method. This device, through joint analysis of organizational data from different business organizations, extracts common and differential features to dynamically determine the business early warning threshold, improving the accuracy and adaptability of early warning threshold setting, effectively reducing false alarm and false negative rates. Simultaneously, by combining business registration and historical behavior information, it dynamically selects appropriate target early warning methods to choose suitable early warning channels for early warning, thereby improving early warning efficiency.

[0067] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 7As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database is used for data employed in the business early warning method. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a business early warning method.

[0068] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described business early warning method.

[0069] In one embodiment, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described business early warning method.

[0070] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0071] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0072] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A business early warning method, characterized in that, include: Based on the customer's identity information and historical behavior information, determine the customer's business score in the business scenario; Based on the business data of multiple different business units corresponding to the business scenario, the common business characteristics and different business characteristics of the business units are determined. Based on the common business characteristics and the different business characteristics, as well as the business organization to which the customer belongs, the business warning threshold for the customer in the business scenario is determined; When the business score exceeds the business warning threshold, the target business warning method for the customer is determined based on the business level of the corresponding business scenario and the historical behavior information, and the target warning information is sent through the target warning method.

2. The method according to claim 1, characterized in that, The step of determining the common and differential business characteristics of the business entities based on organizational business data corresponding to the business scenario includes: Based on the institution's business data, determine the local institution characteristics corresponding to the business scenario; Based on the joint comparison platform, feature alignment is performed on the local institutional features to obtain cross-institutional shared features: Based on the global sharing model, correlation analysis is performed on the cross-organizational shared features to obtain the common business features: The differentiated business characteristics are determined by comparing the local institutional characteristics with the cross-institutional shared characteristics.

3. The method according to claim 1, characterized in that, The step of determining the business warning threshold for the customer in the business scenario based on the common business characteristics, the differentiated business characteristics, and the business organization to which the customer belongs includes: Based on the business organization to which the customer belongs, the common and different business characteristics corresponding to the business scenarios are integrated to obtain integrated business characteristics; Based on the characteristics of the integrated services, the service early warning threshold is determined.

4. The method according to claim 1, characterized in that, Based on the business level corresponding to the business scenario and the historical behavior information, the target business early warning method for the customer is determined as follows: Based on the business level, determine the business value of the business scenario; Based on the historical behavior information, determine the customer's response time among multiple candidate service warning methods; Based on the business value and the response time, a target response method is determined from the multiple candidate warning methods.

5. The method according to claim 1, characterized in that, The sending of target warning information includes: Based on the historical behavior information, obtain the customer's historical warning dialogue information in the business scenario; Based on the target response method and the business scenario, candidate warning information is generated using a natural language model; Based on the historical warning dialogue information, the candidate warning information is modified to generate the target warning information, and the target warning information is sent to the customer.

6. The method according to claim 1, characterized in that, Before determining the customer's business score in the business scenario based on the customer's identity information and historical behavior information, the method further includes: Obtain the customer's business data within a preset time period; Based on the business data, determine whether the customer meets the early warning triggering conditions; When the customer meets the warning triggering conditions, the identity information and the historical behavior information are obtained.

7. The method according to claim 6, characterized in that, After sending the target warning information, the following is also included: Obtain the response operation corresponding to the target warning information; Based on the response operation, the business conversion parameters of the target early warning information are determined through a causal model; Based on the business conversion parameters, the early warning triggering conditions and / or the business early warning threshold are updated.

8. A business early warning device, characterized in that, include: The rating determination module is used to determine the customer's business rating in a business scenario based on the customer's identity information and historical behavior information. The feature determination module is used to determine the common and differential business features of the business entities based on the organizational business data of multiple different business entities corresponding to the business scenario. The threshold determination module is used to determine the business warning threshold of the customer in the business scenario based on the common business characteristics, the different business characteristics, and the business organization to which the customer belongs; The early warning execution module is used to determine the target business early warning method for the customer based on the business level of the corresponding business scenario and the historical behavior information when the business score exceeds the business early warning threshold, and to send the target early warning information through the target early warning method.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the business early warning method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the business early warning method as described in any one of claims 1 to 7.

Images