Service duration determination method and device, equipment and storage medium

Abstract

The invention provides a service duration determination method and device, equipment and a storage medium, relates to the field of big data, and is used for improving the efficiency and rationality of continued service duration determination. According to the specific scheme, the method comprises the steps of obtaining basic information of a target object, intelligently matching a pre-training model in combination with industry and resource features, predicting the total service duration by using a survival analysis method, and finally performing dynamic calibration based on a service score to determine the target service duration for providing resource renewal service for a service object. Automatic and precise determination of the service renewal duration is realized.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the field of big data, and in particular to a service duration determination method and device, equipment and a storage medium. BACKGROUND

[0002] In the field of resource services, when a service object proposes a renewal service of a resource service, a service provider needs to determine a renewal duration for the renewal service of the existing service object. How to balance the continuity of the service relationship and the potential risks, and provide a reasonable renewal duration, is a problem that needs to be solved.

[0003] In the prior art, the service personnel of the service provider usually makes a manual decision according to experience, or determines the renewal duration according to the service duration decision standard corresponding to the first resource service.

[0004] However, the manual method is inefficient and the standards are not uniform, and it is difficult to accurately assess the actual risks of the renewal service according to the decision standard of the first resource service, and it is impossible to quickly and accurately determine a reasonable renewal duration. SUMMARY

[0005] Embodiments of the present application provide a service duration determination method, device, equipment and storage medium, for improving the efficiency and rationality of the renewal service duration determination.

[0006] To achieve the above-mentioned purpose, embodiments of the present application adopt the following technical solutions: In a first aspect, a service duration determination method is provided, the method comprising: obtaining basic information of a target object; wherein the basic information includes a current served duration of the target object having accepted a resource service, industry information to which the target object belongs, and resource information of the target object; determining a target model applicable to the target object from a plurality of preset models based on the industry information and the resource information; the plurality of models are obtained by pre-training service data of a plurality of historical service objects; obtaining resource state feature data of the target object, and predicting a total service duration of the target object from a service starting point of the resource service as a service survival starting point based on the target model and the resource state feature data, to obtain a predicted total service duration; wherein the total service duration represents a continuous duration from the service starting point as a starting time point without a preset termination event; taking a difference between the predicted total service duration and the current served duration as a first recommended service duration for a renewal service of the resource service provided to the target object; calibrating the first recommended service duration according to pre-evaluated service score information of the target object, to obtain a target service duration.

[0007] The embodiment of the application first acquires the current service duration, industry information and resource information of the target object, covers key dimensions such as service duration status, industry attribute and resource information; then according to the industry and resource characteristics, a target model is matched from a plurality of preset models, so as to avoid the prediction deviation of a general model and improve the pertinence and reliability of the total service duration prediction; then the total service duration is predicted through the model output, and the first recommended service duration is calculated by combining the current service duration, so as to preliminarily determine the reference value of the renewal duration; finally, the service score is dynamically calibrated according to the service history, credit status and other dimensions, so as to control the service risk. The method breaks through the limitations of relying on artificial experience, low decision-making efficiency and inaccurate risk control in the traditional renewal duration determination, can quickly output a reasonable renewal suggestion, and can improve the decision-making efficiency of the service provider, accurately control the renewal risk, and at the same time, guarantee the reasonable renewal demand of the target object.

[0008] In a possible implementation manner of the first aspect, the total service duration of the target object from a service starting point of the resource service as a service survival starting point is predicted based on the target model and the resource state feature data to obtain a predicted total service duration, including: According to a first expression, a first function corresponding to the target object is determined; the first expression is: H(t) = h0(t) × exp(b1×x1+b2×x2+…+bk×xk); Wherein, H(t) is the first function, H(t) is used to represent the instantaneous probability of the target object occurring a preset termination event at time t; h0(t) is a baseline function corresponding to the target model that changes with time t, b1, b2…bk are weight coefficients in the target model, x1, x2…xk are resource state feature data; According to a second expression, a service survival curve corresponding to the target object is obtained; the second expression is: S(t) = exp(-∫H(t) dt); Wherein, S(t) represents the service survival curve, S(t) represents the service survival probability of the target object at time point t from the service starting point of the resource service; A target time point corresponding to a preset survival probability in the service survival curve is determined, and an interval duration between the target time point and the service starting point is taken as the predicted total service duration.

[0009] Based on the above technical content, this embodiment determines the first function corresponding to the target object through a first expression, quantifying the impact of various resource status characteristics on service termination risk; obtains the service survival curve through a second expression, intuitively displaying the changing pattern of service continuity probability over time; and determines the predicted total service duration based on the preset survival probability, providing a clear quantitative basis for renewal decisions. This method can scientifically handle uncertainties in the service cycle and improve the accuracy of duration prediction.

[0010] In one possible implementation of the first aspect, the target model corresponds to a confidence prediction interval, which is defined by an upper limit value and a lower limit value; The step of calibrating the first suggested service duration based on the pre-assessed service rating information of the target object to obtain the target service duration includes: Based on the service rating information, a second suggested service duration and a service level corresponding to the target object are determined; the service level is used to characterize the probability of a preset termination event occurring. According to the third expression, the maximum recommended service duration for providing the renewal service to the target object is determined; the third expression is: Ta = [Tr / C] × C; Where Ta represents the maximum suggested service duration, Tr represents the first suggested service duration, C is a preset value, and [ ] indicates rounding down; and the preset maximum service duration and preset minimum service duration corresponding to the resource service are obtained; If the predicted total service duration is within the confidence prediction interval, the maximum suggested service duration is adjusted according to the relationship between the first suggested service duration and the second suggested service duration, as well as the service level, to obtain the target service duration. If the predicted total service duration is greater than the upper limit, and the first suggested service duration is greater than the second suggested service duration, then the minimum value between the first suggested service duration and the preset maximum service duration shall be taken as the target service duration. If the predicted total service duration is less than the lower limit, and the first suggested service duration is less than or equal to the second suggested service duration, then the target service duration is determined from the preset minimum service duration and the first suggested service duration according to the service level.

[0011] Here, this embodiment assesses the reliability of the model's predictions using confidence prediction intervals, determines the second suggested service duration and service level by combining service score information, and then calculates the maximum suggested service duration using a third expression. Based on the relative position of the predicted total service duration within the confidence interval, the duration recommendations are adjusted differentially according to the service level. By implementing corresponding risk controls based on the reliability of the model's prediction results, the final determined target service duration can achieve both reliability and risk controllability.

[0012] In one possible implementation of the first aspect, the service level includes a first level, a second level, a third level, and a fourth level; the probability of the preset termination event occurring at the first level, the second level, the third level, and the fourth level increases sequentially. The step of adjusting the maximum suggested service duration based on the relationship between the first suggested service duration and the second suggested service duration, and the service level, to obtain the target service duration includes: If the first suggested service duration is less than or equal to the second suggested service duration, and the service level corresponding to the target object is the third level, then the difference between the maximum suggested service duration and the preset duration shall be taken as the target service duration. If the first suggested service duration is less than or equal to the second suggested service duration, and the service level corresponding to the target object is the first level, then the sum of the maximum suggested service duration and the preset duration shall be used as the target service duration. If the first suggested service duration is longer than the second suggested service duration, and the service level is the fourth level, then the target service duration is determined to be zero. If the first suggested service duration is greater than the second suggested service duration, and the service level corresponding to the target object is any one of the first level, the second level, or the third level; or if the first suggested service duration is less than or equal to the second suggested service duration, and the service level corresponding to the target object is the second level, then the maximum suggested service duration shall be taken as the target service duration.

[0013] In this embodiment, the relationship between the first and second suggested service durations and the service level are combined to achieve differentiated duration adjustments: service durations are appropriately extended for low-risk levels, while a conservative strategy or even service termination is adopted for high-risk levels. This refined adjustment controls service risk and optimizes the allocation of service resources.

[0014] In one possible implementation of the first aspect, determining the target service duration based on the service level, between the preset minimum service duration and the first suggested service duration, includes: If the service level corresponding to the target object is the third level, then the minimum value between the first suggested service duration and the preset minimum service duration shall be used as the target service duration; If the service level corresponding to the target object is the first level, then the maximum value between the first suggested service duration and the preset minimum service duration shall be used as the target service duration; If the service level corresponding to the target object is the second level, then the preset minimum service duration is taken as the target service duration.

[0015] Based on the above technical content, this embodiment combines service levels with preset minimum service duration and first recommended service duration to achieve intelligent boundary management of service duration: strict minimum duration control is implemented for high-risk levels, while more flexible duration selection is provided for low-risk levels. This differentiated boundary management strategy can ensure service security.

[0016] In one possible implementation of the first aspect, the method further includes: Based on service data from multiple historical service objects, the Breuer score of the model at each prediction time point is calculated; the magnitude of the Breuer score is negatively correlated with the prediction accuracy of the model. The continuous time interval in which the Breuer score remains below a preset threshold is used as the confidence prediction interval for the model.

[0017] Here, the accuracy of model predictions is continuously evaluated through Breuer scores, and prediction stability is used as the basis for determining confidence intervals, providing a reliable and credible reference for service duration calibration and enhancing the scientific nature of service duration decisions.

[0018] In one possible implementation of the first aspect, after obtaining the target service duration, the method further includes: Based on the weight coefficients in the target model, at least one key feature is determined from the resource status feature data; If the weight coefficient corresponding to the key feature is positive, then output the suggestion information to reduce the value corresponding to the key feature; If the weight coefficient corresponding to the key feature is negative, then a suggestion to increase the value of the key feature will be output.

[0019] In this embodiment, key influencing factors are identified from resource status characteristic data by analyzing the weight coefficients in the target model, and specific optimization suggestions are provided based on the positive or negative value of the weight coefficients. This method reveals the key factors affecting service continuity and provides a clear direction for service improvement.

[0020] In one possible implementation of the first aspect, before determining the target model suitable for the target object from a predefined plurality of models based on the industry information and the resource information, the method further includes: Acquire service data for multiple historical service objects; wherein, the service data includes the service start time, service end time, identifier of whether a preset termination event has occurred, industry category information, resource information, and status characteristic data of the service object; Based on the service data of the aforementioned multiple historical service objects, a training sample set and a test sample set are constructed. Based on industry category information and resource information, the training sample set is divided into multiple training subsets and corresponding multiple test subsets; For each training subset and its corresponding test subset, perform the following steps: The historical state feature data of historical service objects in the training subset are used as training input data, and the corresponding service life and whether a preset termination event occurs are used as training labels. Based on the training input data and training labels, a candidate model is trained and obtained; The candidate model is tested using a test subset corresponding to the training subset. If the test results meet the preset performance criteria, the candidate model is determined to be a trained model.

[0021] In this embodiment of the application, the training sample set is divided according to industry category information and resource information, and multiple models are trained respectively. This fully considers the characteristic differences of different types of service objects and significantly improves the prediction accuracy of the model in various sub-scenarios.

[0022] In one possible implementation of the first aspect, the historical state feature data includes historical resource attribute feature data and historical resource change feature data; The construction of a training sample set based on the service data of the multiple historical service objects includes: From a pool of historical service recipients, select the service recipients who first request the renewal of the resource service within a preset first time period as candidate samples; For each candidate sample, perform the following processing to obtain the training samples in the training sample set: The service life start point is the time when the corresponding historical service object first applies for resource services, and the observation time point is the time when the historical service object first renews its service. Extract the historical resource attribute feature data of the historical service object at the observation time point, and extract the historical resource change feature data of the historical service object during the observation period; wherein, the observation period is a preset time period from the observation time point forward; The historical resource attribute feature data and the historical resource change feature data are used as the historical state feature data of the candidate samples; If a preset termination event occurs for the first time before the preset observation deadline, the candidate sample is marked with an event occurrence sample identifier, and the service survival time of the candidate sample from the service survival start point to the event occurrence is recorded. Otherwise, the candidate sample is marked with an event non-occurrence sample identifier, and the service survival time of the candidate sample from the service survival start point to the observation deadline is recorded.

[0023] Here, this embodiment constructs training samples that can accurately reflect the service lifecycle and meet the training requirements of the survival analysis model by taking the time of the first resource service application as the service survival start point, the time of the first service renewal as the observation point, and marking samples and recording service survival duration based on whether a preset termination event occurs, thus laying the foundation for obtaining an accurate prediction model.

[0024] Secondly, a service duration determination device is provided, the device comprising: An acquisition unit is used to acquire basic information about a target object; wherein, the basic information includes the current service duration for which the target object has received resource services, the industry information to which the target object belongs, and the resource information of the target object; The first processing unit is used to determine a target model suitable for the target object from a set of preset models based on the industry information and the resource information; the set of preset models are pre-trained from service data of multiple historical service objects. The second processing unit is used to acquire resource status feature data of the target object, and based on the target model and the resource status feature data, to predict the total service life of the target object with the service start point of the resource service as the service life start point, and to obtain the predicted total service life; wherein, the total service life represents the duration during which no preset termination event has occurred with the service start point as the start time point. The determining unit is configured to use the difference between the predicted total service duration and the current service duration as the first suggested service duration for providing the resource service renewal service to the target object; The calibration unit is used to calibrate the first suggested service duration based on the service score information of the target object that has been pre-evaluated, so as to obtain the target service duration.

[0025] Thirdly, an electronic device is provided, comprising: a memory and at least one processor. The memory is communicatively connected to the processor. The memory is used to store computer program code, which includes computer instructions. When the processor executes the computer instructions, it causes the electronic device to perform the methods described in the first aspect and any possible implementation thereof.

[0026] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer instructions. When executed by a processor, these computer instructions are used to implement the method described in the first aspect and any possible implementation thereof.

[0027] Fifthly, embodiments of this application provide a computer program product that, when run on a computer or executed by a computer's processor, implements the method described in the first aspect and any possible design thereof. The computer may be the electronic device described in the third aspect and any possible implementation thereof.

[0028] It is understood that the beneficial effects achieved by the service duration determination device described in the second aspect, the electronic device described in the third aspect, the computer-readable storage medium described in the fourth aspect, and the computer program product described in the fifth aspect can be referred to as the beneficial effects in the first aspect and any possible implementation thereof, and will not be repeated here. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating a method for determining service duration provided in an embodiment of this application; Figure 2 This is a flowchart illustrating another method for determining service duration provided in an embodiment of this application; Figure 3 This is a schematic diagram of a service duration determination device provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0030] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this embodiment, unless otherwise stated, "a plurality of" means two or more.

[0031] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0032] The technical solutions provided in this application, including the collection, storage, use, processing, transmission, provision, and disclosure of financial data or user data, comply with relevant laws and regulations and do not violate public order and good morals.

[0033] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0034] In related technologies, when determining the renewal duration of services for service recipients, business personnel usually make manual decisions based on experience, or directly use the service duration decision rules corresponding to all or part of the initial resource services to determine the renewal duration. However, manual methods are inefficient and inconsistent in standards. It is difficult to accurately assess the actual risks of renewal services based on the decision standards of initial resource services, and neither method can quickly and accurately determine a reasonable renewal duration.

[0035] To improve the efficiency and rationality of determining the renewal service duration, the embodiments of this application obtain the basic information of the target object, combine industry and resource characteristics to intelligently match the pre-trained model, use survival analysis methods to predict the total service duration, and perform dynamic calibration based on service scores, thereby realizing the automated and accurate determination of the renewal service duration.

[0036] This application provides a method for determining service duration, which can be applied to electronic devices. The electronic device can be a single server, a server cluster consisting of multiple servers, a cloud computing platform with data processing capabilities, an edge computing device, a chip, or a device with computing capabilities. This application does not limit the specific form of the electronic device.

[0037] Figure 1 This is a flowchart illustrating a method for determining service duration provided in an embodiment of this application. Figure 1 As shown, the method in the embodiments of this application may include: S101. Obtain basic information about the target object; wherein, the basic information includes the current service duration of the resource services received by the target object, the industry information of the target object, and the resource information of the target object.

[0038] The target object can refer to an entity that has received resource services (such as financial services, cloud resource services, energy services, and digital resource services), such as a company or individual; the current service duration is the duration from the start point of the resource service to the current time; industry information is used to characterize the industry category to which the target object belongs (such as manufacturing, information technology, etc.); resource information can include resource holding type, resource holding quantity, etc.

[0039] Service duration can be measured in years, months, or days; this example uses months for explanation.

[0040] For example, this embodiment can receive basic information about the target object through an application programming interface (API) with the resource service management system or through a user interface (such as a web form or client input box). For instance, the user can input the target object's current service duration as "12 months", industry information as "finance", and resource information as "loan amount of 1 million".

[0041] S102. Based on industry information and resource information, determine the target model suitable for the target object from multiple preset models; the multiple models are pre-trained from service data of multiple historical service objects.

[0042] For example, this embodiment pre-stores multiple models, each trained based on historical service data of service objects from different industries and resource types, which can adapt to the service duration prediction needs in different scenarios. Historical service objects refer to entities that have historically received resource services or resource service renewal services. Their service data includes service start time, service end time, identifiers of whether a preset termination event (such as service default, resource termination, etc.) has occurred, industry category information, resource information, and status characteristic data, etc.

[0043] In this embodiment, based on the industry and resource information of the target object, the most suitable target model is selected from the preset model pool through rule matching or similarity calculation.

[0044] S103. Obtain resource status feature data of the target object. Based on the target model and resource status feature data, predict the total service life of the target object with the service start point of resource service as the service life start point, and obtain the predicted total service life. The total service life represents the duration without a preset termination event occurring from the service start point as the starting time point.

[0045] The resource status characteristic data may include the dynamic characteristics of the target object during the process of receiving resource services and the time characteristics when applying for resource renewal services, such as resource holding status, service performance records, and operational stability indicators; the total service life refers to the duration from the service start point to the first occurrence of a preset termination event (such as service interruption, default triggering, etc.), which is used to quantify the expected duration of the service relationship.

[0046] For example, in this embodiment, the target model is invoked, resource status feature data is input, and the model predicts the total service duration. For instance, if the target model outputs a predicted total service duration of "60 months," it means that the target object is expected to continuously receive resource services for 60 months from the service start point without any termination event.

[0047] S104. The difference between the predicted total service duration and the current service duration is used as the first suggested service duration for the renewal service of the resource service provided to the target object.

[0048] For example, in this embodiment, the difference between the predicted total service duration and the current service duration is calculated to obtain a first suggested service duration, which represents the serviceable duration of the resource renewal service based on the model prediction.

[0049] For example, if the predicted total service duration is 60 months and the current service duration is 12 months, then the first recommended service duration is 48 months.

[0050] S105. Based on the service rating information of the target object in the pre-assessment, the first suggested service duration is calibrated to obtain the target service duration.

[0051] Among them, the service rating information can be a quantitative score that is pre-assessed based on the target object's service history, credit status, industry risk and other dimensions, and is used to reflect the service sustainability and risk level.

[0052] For example, this embodiment can dynamically adjust the first recommended service duration based on service rating information: if the service rating is high, it indicates strong service stability and low risk, and the first recommended service duration can be appropriately extended; if the service rating is low, the first recommended service duration is shortened to control risk. The target service duration is obtained after calibration and serves as the final determined renewal service duration. For example, if the first recommended service duration is 48 months, after service rating calibration, the target service duration is adjusted to 36 months.

[0053] In summary, this embodiment first obtains the current service duration, industry information, and resource information of the target object, covering key dimensions such as service availability, industry attributes, and resource information. Then, based on industry and resource characteristics, it matches a suitable target model from multiple preset models to avoid prediction biases in general models and improve the targeting and reliability of the total service duration prediction. Next, it calculates the first suggested service duration by combining the predicted total service duration output by the model with the current service duration, thus initially clarifying the reference value for the renewal duration. Finally, it dynamically calibrates based on service scores from dimensions such as service history and credit status to control service risks. This approach overcomes the limitations of traditional renewal duration determination, which relies on human experience, has low decision-making efficiency, and inaccurate risk control. It can quickly output reasonable renewal suggestions, improving the service provider's decision-making efficiency, accurately controlling renewal risks, and ensuring the target object's reasonable renewal needs.

[0054] Figure 2 This is a flowchart illustrating another method for determining service duration provided in an embodiment of this application. Figure 2 As shown, the method in the embodiments of this application may include: S201. Obtain basic information about the target object; wherein, the basic information includes the current service duration of the resource services received by the target object, the industry information of the target object, and the resource information of the target object.

[0055] The target audience refers to entities that have accepted resource services and applied for renewal. In the resource service scenario, these can be enterprises, institutions, or individual users. Taking bank credit services as an example, the target audience is customers who apply for loan renewal.

[0056] Taking bank credit services as an example, industry information is used to identify the category of economic activity to which the target object belongs; resource information comprehensively reflects the resource load and structural risks of the target object. For example, in a credit scenario, it may include asset and liability status, debt scale, guarantee status, etc.

[0057] For example, this embodiment obtains basic information about the target object through a system interface or interactive interface. For instance, this embodiment obtains that the target object's current service duration is "18 months", its industry information is "manufacturing", and its resource information is "asset-liability ratio of 60%".

[0058] S202. Based on industry information and resource information, determine the target model suitable for the target object from multiple preset models; the multiple models are pre-trained from service data of multiple historical service objects.

[0059] For example, in this embodiment, based on the industry information and resource information of the target object, a suitable target model is selected from a preset model pool through rule matching or similarity calculation.

[0060] Here, "historical service recipients" refers to entities that have historically received resource services; in the banking credit scenario, historical service recipients are customers who have previously applied for or renewed loans. Multiple models are trained based on historical data from different industries and resource types, which can improve the targeting of predictions.

[0061] In one example, prior to S202, the embodiments of this application further include: Acquire service data for multiple historical service objects; the service data includes the service start time, service end time, identifier of whether a preset termination event has occurred, industry category information, resource information, and status characteristic data of the service object; Based on service data from multiple historical service objects, a training sample set and a test sample set are constructed. Based on industry category information and resource information, the training sample set is divided into multiple training subsets and corresponding multiple test subsets; For each training subset and its corresponding test subset, perform the following steps: The historical state feature data of historical service objects in the training subset are used as training input data, and the corresponding service life and whether a preset termination event occurs are used as training labels. Based on the training input data and training labels, a candidate model is trained to obtain the model. The candidate model is tested using the test subset corresponding to the training subset. If the test results meet the preset performance criteria, the candidate model is determined to be a well-trained model.

[0062] In one feasible implementation, the historical state feature data includes historical resource attribute feature data and historical resource change feature data; based on the service data of multiple historical service objects, a training sample set is constructed, including: From multiple historical service recipients, those who first requested resource service renewal within a preset first time period were selected as candidate samples. For each candidate sample, perform the following processing to obtain the training samples in the training sample set: The service survival start point is the time when the corresponding historical service object first applies for resource services, and the observation point is the time when the historical service object first renews its services. Extract historical resource attribute feature data of historical service objects at the observation time point, and extract historical resource change feature data of historical service objects during the observation period; wherein, the observation period is a preset time period from the observation time point forward; Historical resource attribute feature data and historical resource change feature data are used as historical state feature data for candidate samples; If a preset termination event occurs for the first time before the preset observation deadline, the candidate sample is marked with the event occurrence sample identifier, and the service survival time of the candidate sample from the service survival start point to the event occurrence is recorded. Otherwise, the candidate sample is marked with the event non-occurrence sample identifier, and the service survival time of the candidate sample from the service survival start point to the observation deadline is recorded.

[0063] In one feasible implementation, the target model corresponds to a confidence prediction interval, which is defined by an upper limit and a lower limit; embodiments of this application also include: Based on service data from multiple historical service objects, the Breuer score of the model at each prediction time point is calculated; the magnitude of the Breuer score is negatively correlated with the prediction accuracy of the model. The continuous time interval in which the Breuer score remains below a preset threshold is used as the confidence prediction interval for the model.

[0064] For example, the process of pre-training multiple models in this embodiment is as follows: Service data from multiple historical loan renewal customers was extracted from the bank's credit system, including: Basic service information: start and end times of loan services for historical customers; Event identifier: An identifier indicating whether a preset termination event (e.g., 30+ days overdue) has occurred (1 = occurred, 0 = did not occur). For example, if a customer's loan becomes 30+ days overdue during the service life, the preset termination event is considered to have occurred. If the customer settles the loan normally or early, the event is considered not to have occurred. Category dimension information: industry category of historical customers (such as manufacturing, service industry), resource information (such as debt-to-asset ratio of 50%); Status characteristic data: historical customer's asset and liability data when applying for loan renewal, such as customer basic information, customer liability information, customer asset information, customer collateral information, customer credit information, and resource change characteristic data for the preset time period before applying for loan renewal, such as business change information, loan usage change information, customer interaction information with our bank, asset change information, credit change information, liability change information, and basic information change status.

[0065] Building training and testing sets: Sample construction: If multiple loans for the same customer are separated by ≥M months (e.g., M is set to 1~3), they are considered as 2 independent samples; the earliest loan application time of the customer is taken as the "service survival start point"; only the customer's first loan renewal data is retained for modeling; the sample may be right-censored (i.e., no preset termination event is observed), and there is no left-censoring.

[0066] Samples of first-time loan renewals within a preset historical time period were selected and randomly divided into training and test sets in a 7:3 ratio. Samples of first-time loan renewals within the preset historical time period were used as an "out-of-time validation set" to avoid sample time bias.

[0067] Divide the training / test subsets: Divide the training set and test set based on "industry category + debt-to-asset ratio" to obtain i training subsets and i corresponding test subsets.

[0068] Training candidate models: For each training subset, the training input is "historical customer status feature data", and the training labels are "service survival time" (number of months from the start of survival to the first 30+ overdue period) and "whether a 30+ overdue period has occurred". Construct a COX model as a candidate model. Each candidate model contains k input features (each model may use the same or different input features).

[0069] Testing and Model Determination: In this embodiment, candidate models are tested using corresponding test subsets, and i models are finally obtained, which constitute multiple preset models.

[0070] This embodiment trains models by classifying them according to industry and resource information, making each model suitable for a specific customer group. This avoids the prediction bias of general models for customer groups with different risk characteristics, effectively improving the targeting of subsequent target model selection and improving prediction accuracy, thereby effectively balancing the continuity of service relationships with potential risks.

[0071] When constructing samples, this embodiment takes the first resource service application time of the candidate sample as the service survival start point; the first service renewal time of the candidate sample as the observation time point; and a preset time period before the observation time point as the observation period.

[0072] Historical resource attribute characteristic data is point-in-time characteristic data: including, for example, static data at the observation point in time, such as total assets, total liabilities, quantity of collateral, credit score, etc. Historical resource change characteristic data are statistical characteristic data, including dynamic data during the observation period, such as the monthly average growth rate of loan amount, the change in the asset-liability ratio, and the average number of monthly interactions with banks.

[0073] Set a preset observation deadline. If a candidate sample first experiences a preset termination event (such as 30+ overdue) before the preset observation deadline, it is marked as an event occurrence sample with a corresponding label of 1, and the service survival time is recorded, for example, the number of months from the start of survival to the occurrence of the event. If a candidate sample does not experience a preset termination event before the preset observation deadline (e.g., normal settlement), it is marked as a sample where the event did not occur, with a corresponding label of 0, and the corresponding service lifetime is recorded.

[0074] This embodiment ensures that the feature data of the training samples are highly matched with the actual loan renewal scenarios by clearly defining the time benchmark and feature classification; at the same time, it accurately portrays the risk outcomes of historical customers by defining the "event occurred / did not occur" label, providing high-quality samples for model training.

[0075] Furthermore, in this embodiment, the time period in which the prediction results of the model are better is determined by using the Brier score (BS) equal to a preset threshold, such as BS=0.2, and is denoted as the confidence prediction interval [Tbi, Tei].

[0076]

[0077] Where N is the number of samples, ft is the model's predicted probability for the i-th sample, and ot is the actual result for the i-th sample (1 = event occurred, 0 = event did not occur); the smaller the Brier Score, the higher the model's prediction accuracy at that time point.

[0078] This embodiment quantifies the accuracy of the model at different time points using the Breuer score, thereby determining the confidence prediction interval. This provides a basis for subsequent judgment on whether the predicted total service duration is reliable. It can avoid errors in renewal decisions caused by using time ranges with poor model prediction performance, thus effectively balancing the continuity of service relationships with potential risks.

[0079] S203. Obtain the resource status characteristic data of the target object.

[0080] Among them, resource status characteristic data is comprehensive data that characterizes the resource attributes and change characteristics of the target object during the resource service renewal application stage, and can include static point-in-time characteristic data and dynamic change characteristic data.

[0081] Taking bank credit services as an example, static point-in-time characteristic data can correspond to the customer's basic information, liability information, asset information, collateral information, credit information, etc. when the customer applies for loan renewal; dynamic change characteristic data can correspond to the customer's business changes, loan usage changes, interaction information with the bank, asset changes, credit changes, liability changes, etc. during the observation period (such as 1 year before the loan renewal application).

[0082] For example, this embodiment obtains resource status characteristic data of the target object through channels such as the bank's core business system interface and credit management system. For example, for manufacturing customer A applying for loan renewal, static time-point characteristic data at the time of application is extracted: the company has been established for 5 years, the asset-liability ratio is 60%, there is 1 piece of real estate as collateral, and there is no overdue credit record; at the same time, dynamic change characteristic data over the past year is extracted: the average monthly loan amount increased by 10% compared with the previous year, the average number of monthly interactions with the bank is 8, the current assets increased by 5% compared with the beginning of the year, and there is no record of new liabilities.

[0083] For different target models, the categories of resource state feature data to be acquired may be the same or different, corresponding to the categories of historical state feature data in the training data of the model.

[0084] This embodiment comprehensively acquires resource status characteristic data combining static and dynamic data, providing complete data support for the accurate prediction of the target model, and ensuring that the prediction results reflect the true resource status and development trend of the target object.

[0085] S204. Based on the target model and resource status feature data, predict the total service life of the target object with the service start point of resource service as the service life start point, and obtain the predicted total service life; wherein, the total service life represents the duration without the occurrence of the preset termination event with the service start point as the start time point.

[0086] In one feasible implementation, S204 includes the following steps: Based on the first expression, determine the first function corresponding to the target object; the first expression is: H(t) = h0(t) × exp(b1×x1+b2×x2+…+bk×xk); Wherein, H(t) is the first function, which is used to characterize the instantaneous probability of the target object experiencing a preset termination event at time t; h0(t) is the benchmark function corresponding to the target model that changes with time t; b1, b2...bk are the weight coefficients in the target model; and x1, x2...xk are the resource status feature data. According to the second expression, the service survival curve corresponding to the target object is obtained; the second expression is: S(t) = exp(-∫H(t) dt); Wherein, S(t) represents the service survival curve, and S(t) represents the service survival probability of the target object at time point t, calculated from the service start point of the resource service; Determine the target time point in the service survival curve that corresponds to the preset survival probability, and use the interval between the target time point and the service start point as the predicted total service duration.

[0087] For example, this embodiment predicts the total service lifetime based on the target model and resource status feature data through the following steps: The first function corresponding to the target object is determined according to the first expression, which is the Cox model, denoted as: H(t)=h0(t)×exp(b1×x1+b2×x2+…+bk×xk)(1) Wherein, H(t) is the first function, representing the instantaneous probability of the target object experiencing a preset termination event at time t. Taking a bank credit scenario as an example, the preset termination event is a loan overdue of 30+ months, and H(t) is the instantaneous risk probability of the customer experiencing a 30+ month overdue period in month t; h0(t) is the benchmark function corresponding to the target model that changes with time t, i.e., the benchmark risk rate function, representing the probability of the target object experiencing a preset termination event at time t when there are no resource status features; b1, b2...bk are the weight coefficients in the target model, obtained from historical samples, used to quantify the degree of influence of each resource status feature on the probability of the preset termination event; x1, x2...xk are resource status feature data, i.e., the resource attributes and change feature data of the target object, which can include static point-in-time feature data and dynamic change feature data.

[0088] The service survival curve corresponding to the target object is obtained according to the second expression, which is: S(t)=exp(-∫H(t)dt)(2) S(t) represents the service survival curve, and its core meaning is the service survival probability of the target object at different time points t, calculated from the service start point of the resource service, if the preset termination event has not occurred. ∫H(t)dt is the integral of the first function H(t) from 0 to t, that is, the cumulative risk rate. The larger the integral result, the lower the survival probability.

[0089] This embodiment determines the target time point on the service survival curve corresponding to the preset survival probability, and uses the interval between the target time point and the service start point as the predicted total service duration. Taking a bank loan scenario as an example, the preset survival probability is 80%, that is, finding the time point t=36 months corresponding to S(t)=80% on the service survival curve indicates that the customer has an 80% probability of not incurring a 30+ month delinquency within 36 months from the loan start point. Therefore, the predicted total service duration is 36 months.

[0090] This embodiment uses a quantitative function model and survival curve analysis to directly correlate resource status characteristics with survival probability, making the calculation of predicted total service time more scientific and accurate, and avoiding the bias of subjective experience judgment.

[0091] S205. Take the difference between the predicted total service duration and the currently served duration as the first recommended service duration for the renewal service of providing resource services to the target object.

[0092] Taking the bank loan renewal scenario as an example, in this embodiment, the current resource service of the customer in the bank is obtained in advance. For example, the remaining duration of the loan is denoted as Tn, the calculated predicted total service duration is denoted as Mi, and the maximum loan renewal time of the customer is denoted as the first recommended service duration Tr for providing the loan renewal service to the customer, and we get: Tr = Mi - Tn (3) S206. Determine the second recommended service duration for the target object and the service level corresponding to the target object according to the pre - evaluated service score information of the target object; where the service level is used to characterize the possibility of a preset termination event.

[0093] Taking the bank loan renewal scenario as an example, the service score information can refer to the quantified score S pre - evaluated based on factors such as the historical service records and resource status of the target object, which is used to preliminarily judge the renewal risk of the target object; the second recommended service duration is the default renewal duration reference value Tf determined based on this score information; the service level is the risk level divided according to the service score information S, which is used to characterize the possibility of a preset termination event for the target object.

[0094] In one example, the service level is divided according to the score card score S: S is greater than the third threshold, the service level is extremely high risk; S is greater than the second threshold, the service level is high risk; the first threshold < S < the second threshold, the service level is medium risk; S is less than the first threshold, the service level is low risk; the first threshold, the second threshold, the third threshold, and the fourth threshold increase in sequence.

[0095] In this embodiment, the service level is quickly divided and the second recommended service duration is determined through the service score information, providing a basic risk reference for the subsequent correction of the target service duration, making the renewal duration recommendation more in line with the risk profile of the target object.

[0096] S207. Determine the maximum recommended service duration for providing the renewal service to the target object according to the third expression; the third expression is: Ta = [Tr / C] × C; where Ta represents the maximum recommended service duration, Tr represents the first recommended service duration, C is a preset value, and [] represents rounding down; and obtain the preset maximum service duration and the preset minimum service duration corresponding to the resource service.

[0097] Exemplarily, in this embodiment, according to the service duration boundary values preset for the resource service, constraint conditions are provided for subsequent calibration.

[0098] S208. If the predicted total service duration is within the confidence prediction range, the maximum suggested service duration is adjusted according to the relationship between the first suggested service duration and the second suggested service duration, as well as the service level, to obtain the target service duration.

[0099] In one example, the service levels include Level 1, Level 2, Level 3, and Level 4; the probability of a preset termination event occurring increases sequentially from Level 1 to Level 4. S208 includes the following steps: If the first suggested service duration is less than or equal to the second suggested service duration, and the service level corresponding to the target object is the third level, then the difference between the maximum suggested service duration and the preset duration will be used as the target service duration. If the first suggested service duration is less than or equal to the second suggested service duration, and the service level corresponding to the target object is the first level, then the sum of the maximum suggested service duration and the preset duration will be used as the target service duration. If the first suggested service duration is greater than the second suggested service duration, and the service level is level four, then the target service duration is determined to be zero. If the first suggested service duration is greater than the second suggested service duration, and the service level corresponding to the target object is any one of the first, second, or third levels; or if the first suggested service duration is less than or equal to the second suggested service duration, and the service level corresponding to the target object is the second level, then the maximum suggested service duration shall be used as the target service duration.

[0100] For example, if the predicted total service duration Mi is within the confidence prediction interval [Tbi, Tei], taking a preset duration of 3 months as an example and C=6 as an example, this embodiment determines the target service duration according to the following steps: If the first recommended service duration Tr = 12 months, then Ta = [Tr / C] × C = 12 months; if the second recommended service duration Tf = 13 months (Tr ≤ Tf), and the customer's corresponding service level is Level 3 (high risk), then the target service duration = Ta - 3 = 9 months; If Tr = 12 months and Tf = 13 months (Tr ≤ Tf), the customer's service level is Level 1 (low risk). In this case, Ta = [Tr / C] × C = 12 months, and the target service duration = 12 + 3 = 15 months. If Tr = 18 months and Tf = 13 months (Tr > Tf), the customer's service level is Level 4 (extremely high risk), then the target service duration = 0 months (reject loan renewal). If Tr = 18 months and Tf = 13 months (Tr > Tf), the customer's service level is Level 2 (medium risk). In this case, Ta = [Tr / C] × C = 18 months, and the target service duration = 18 months. If Tr = 12 months and Tf = 13 months (Tr ≤ Tf), the customer's service level is Level 2 (medium risk). In this case, Ta = [Tr / C] × C = 12 months, and the target service duration is 12 months.

[0101] This embodiment combines the relationship between the first and second suggested service durations and the service level to make differentiated adjustments to the maximum suggested service duration. This takes into account both the prediction results of survival analysis and the scoring card, and can accurately control the risk of loan renewal based on the risk level.

[0102] S209. If the predicted total service duration is greater than the upper limit and the first suggested service duration is greater than the second suggested service duration, then the minimum value between the first suggested service duration and the preset maximum service duration shall be taken as the target service duration.

[0103] For example, the confidence prediction interval for the total service duration is defined by an upper limit value Tei and a lower limit value Tbi (e.g., Tei = 48 months, Tbi = 12 months). The preset maximum service duration is the maximum renewal period stipulated by the policy or product corresponding to the resource service. Taking a preset maximum service duration of 36 months as an example: In this embodiment, if the customer's predicted total service duration is 50 months (greater than the upper limit of 48 months), the current service duration is 18 months, the first suggested service duration Tr = 50 - 18 = 32 months, the second suggested service duration Tf = 13 months, Tr > Tf, then the target service duration = MIN(32, 36) = 32 months.

[0104] In this embodiment, when the predicted total service duration exceeds the model confidence interval and the first suggested service duration is relatively long, the upper limit of the renewal duration is limited by setting a maximum service duration to avoid the risk of excessive loan renewal due to the decline in model prediction performance, while also meeting the compliance requirements of the product service.

[0105] S210. If the predicted total service duration is less than the lower limit, and the first suggested service duration is less than or equal to the second suggested service duration, then the target service duration is determined from the preset minimum service duration and the first suggested service duration according to the service level.

[0106] In one example, S210 includes the following steps: If the service level corresponding to the target object is the third level, then the minimum value between the first suggested service duration and the preset minimum service duration will be used as the target service duration. If the service level corresponding to the target object is the first level, then the maximum value between the first suggested service duration and the preset minimum service duration shall be used as the target service duration; If the service level corresponding to the target object is Level 2, then the preset minimum service duration will be used as the target service duration.

[0107] For example, the preset minimum service duration is the minimum renewal period stipulated by the policy or product corresponding to the resource service. The following explanation uses a preset minimum service duration of 6 months as an example: This embodiment determines the target service duration according to the following steps: If the customer's predicted total service duration is 11 months, which is less than the lower limit of the confidence prediction interval of 12 months, and the service level is high risk with Tf of 13 months, and the current service duration is 4 months, then Tr = 11 - 4 = 7 months. Since Tr ≤ Tf, the target service duration = MIN(7, 6) = 6 months. That is, the preset minimum service duration is the target service duration. If the customer's predicted total service duration is 11 months, which is less than the lower limit of the confidence prediction interval of 12 months, and the service level is low risk with Tf of 13 months, and the current service duration is 4 months, then Tr = 11 - 4 = 7 months. Since Tr ≤ Tf, the target service duration = max(7, 6) = 7 months. That is, the first suggested service duration is the target service duration. If the customer's predicted total service duration is 11 months, which is less than the lower limit of the confidence prediction interval of 12 months, and the service level is medium risk with Tf of 13 months, and the current service duration is 4 months, then Tr = 11 - 4 = 7 months. If Tr ≤ Tf, then the target service duration is preset to a minimum service duration.

[0108] In this embodiment, when the predicted total service duration is lower than the lower limit of the model's confidence interval, the model prediction can be considered relatively conservative. At this time, the service duration of the renewal service is determined by combining the service level and the preset minimum service duration, which realizes differentiated renewal duration control based on risk level and service characteristics. This can avoid low-risk customers being overly restricted, while avoiding high-risk customers being overly relaxed.

[0109] S211. Based on the weight coefficients in the target model, determine at least one key feature from the resource status feature data.

[0110] For example, the weight coefficients b1, b2...bk in the target model quantify the degree of influence of each resource state feature on the probability of the occurrence of the preset termination event. The key feature refers to the resource state feature whose absolute value of the weight coefficient is greater than the preset weight threshold and has a significant impact on the prediction result.

[0111] S212. If the weight coefficient corresponding to the key feature is positive, output the suggestion information to reduce the value of the key feature; if the weight coefficient corresponding to the key feature is negative, output the suggestion information to increase the value of the key feature.

[0112] For example, this embodiment generates targeted optimization suggestions based on the positive and negative relationships of the key feature weights. For features that positively impact risk, it is recommended to decrease their values; for features that negatively impact risk, it is recommended to increase their values.

[0113] This embodiment generates targeted suggestions based on the positive and negative values ​​of the weight coefficients corresponding to key features, enabling service recipients to clearly understand the direction of resource allocation adjustments and facilitate their improvement of qualifications to extend the renewal service period; at the same time, it also helps service providers to prevent and control risks in advance.

[0114] Figure 3 This is a schematic diagram of a service duration determination device provided in an embodiment of this application. Figure 3 As shown, the service duration determination device includes an acquisition unit 301, a first processing unit 302, a second processing unit 303, a determination unit 304, and a calibration unit 305.

[0115] The acquisition unit 301 is used to acquire basic information of the target object; wherein, the basic information includes the current service duration of the resource services received by the target object, the industry information of the target object, and the resource information of the target object; The first processing unit 302 is used to determine a target model suitable for the target object from a set of preset models based on industry information and resource information; the multiple models are pre-trained from service data of multiple historical service objects. The second processing unit 303 is used to acquire resource status feature data of the target object, and based on the target model and resource status feature data, to predict the total service life of the target object with the service start point of the resource service as the service life start point, and to obtain the predicted total service life; wherein, the total service life represents the duration during which no preset termination event occurs with the service start point as the start time point. The determining unit 304 is used to take the difference between the predicted total service duration and the current service duration as the first suggested service duration for the renewal service of providing resource services to the target object; The calibration unit 305 is used to calibrate the first suggested service duration based on the service score information of the target object that has been pre-assessed, so as to obtain the target service duration.

[0116] In other embodiments, the first processing unit 302 is specifically used for: Based on the first expression, determine the first function corresponding to the target object; the first expression is: H(t) = h0(t) × exp(b1×x1+b2×x2+…+bk×xk); Wherein, H(t) is the first function, which is used to characterize the instantaneous probability of the target object experiencing a preset termination event at time t; h0(t) is the benchmark function corresponding to the target model that changes with time t; b1, b2...bk are the weight coefficients in the target model; and x1, x2...xk are the resource status feature data. According to the second expression, the service survival curve corresponding to the target object is obtained; the second expression is: S(t) = exp(-∫H(t) dt); Wherein, S(t) represents the service survival curve, and S(t) represents the service survival probability of the target object at time point t, calculated from the service start point of the resource service; Determine the target time point in the service survival curve that corresponds to the preset survival probability, and use the interval between the target time point and the service start point as the predicted total service duration.

[0117] In other embodiments, the target model corresponds to a confidence prediction interval, which is defined by an upper limit and a lower limit; the calibration unit 305 is specifically used for: Based on the service rating information, a second recommended service duration and a corresponding service level for the target object are determined; the service level is used to characterize the likelihood of a preset termination event occurring. Based on the third expression, determine the maximum recommended service duration for providing renewal services to the target object; the third expression is: Ta = [Tr / C] × C; Where Ta represents the maximum suggested service duration, Tr represents the first suggested service duration, C is the preset value, and [ ] indicates rounding down; and obtain the preset maximum service duration and preset minimum service duration corresponding to the resource service; If the predicted total service duration is within the confidence prediction range, the maximum suggested service duration is adjusted according to the relationship between the first suggested service duration and the second suggested service duration, as well as the service level, to obtain the target service duration. If the predicted total service duration is greater than the upper limit, and the first suggested service duration is greater than the second suggested service duration, then the minimum value between the first suggested service duration and the preset maximum service duration will be used as the target service duration. If the predicted total service duration is less than the lower limit, and the first suggested service duration is less than or equal to the second suggested service duration, then the target service duration is determined from the preset minimum service duration and the first suggested service duration based on the service level.

[0118] In other embodiments, the service levels include a first level, a second level, a third level, and a fourth level; the probability of a preset termination event occurring at the first level, second level, third level, and fourth level increases sequentially; the calibration unit 305 is specifically used for: If the first suggested service duration is less than or equal to the second suggested service duration, and the service level corresponding to the target object is the third level, then the difference between the maximum suggested service duration and the preset duration will be used as the target service duration. If the first suggested service duration is less than or equal to the second suggested service duration, and the service level corresponding to the target object is the first level, then the sum of the maximum suggested service duration and the preset duration will be used as the target service duration. If the first suggested service duration is greater than the second suggested service duration, and the service level is level four, then the target service duration is determined to be zero. If the first suggested service duration is greater than the second suggested service duration, and the service level corresponding to the target object is any one of the first, second, or third levels; or if the first suggested service duration is less than or equal to the second suggested service duration, and the service level corresponding to the target object is the second level, then the maximum suggested service duration shall be used as the target service duration.

[0119] In other embodiments, the calibration unit 305 is further configured to: If the service level corresponding to the target object is the third level, then the minimum value between the first suggested service duration and the preset minimum service duration will be used as the target service duration. If the service level corresponding to the target object is the first level, then the maximum value between the first suggested service duration and the preset minimum service duration shall be used as the target service duration; If the service level corresponding to the target object is Level 2, then the preset minimum service duration will be used as the target service duration.

[0120] In other embodiments, the apparatus further includes a third processing unit for: Based on service data from multiple historical service objects, the Breuer score of the model at each prediction time point is calculated; the magnitude of the Breuer score is negatively correlated with the prediction accuracy of the model. The continuous time interval in which the Breuer score remains below a preset threshold is used as the confidence prediction interval for the model.

[0121] In other embodiments, after calibration unit 305, the device further includes an output unit for: Based on the weight coefficients in the target model, at least one key feature is determined from the resource status feature data; If the weight coefficient corresponding to the key feature is positive, then output the suggestion information to reduce the value of the key feature; If the weight coefficient corresponding to the key feature is negative, then the output will suggest increasing the value of the key feature.

[0122] In other embodiments, prior to the first processing unit 302, the apparatus further includes a training unit for: Acquire service data for multiple historical service objects; the service data includes the service start time, service end time, identifier of whether a preset termination event has occurred, industry category information, resource information, and status characteristic data of the service object; Based on service data from multiple historical service objects, a training sample set and a test sample set are constructed. Based on industry category information and resource information, the training sample set is divided into multiple training subsets and corresponding multiple test subsets; For each training subset and its corresponding test subset, perform the following steps: The historical state feature data of historical service objects in the training subset are used as training input data, and the corresponding service life and whether a preset termination event occurs are used as training labels. Based on the training input data and training labels, a candidate model is trained to obtain the model. The candidate model is tested using the test subset corresponding to the training subset. If the test results meet the preset performance criteria, the candidate model is determined to be a well-trained model.

[0123] In some embodiments, historical state feature data includes historical resource attribute feature data and historical resource change feature data; the training unit is specifically used for: From multiple historical service recipients, those who first requested resource service renewal within a preset first time period were selected as candidate samples. For each candidate sample, perform the following processing to obtain the training samples in the training sample set: The service survival start point is the time when the corresponding historical service object first applies for resource services, and the observation point is the time when the historical service object first renews its services. Extract historical resource attribute feature data of historical service objects at the observation time point, and extract historical resource change feature data of historical service objects during the observation period; wherein, the observation period is a preset time period from the observation time point forward; Historical resource attribute feature data and historical resource change feature data are used as historical state feature data for candidate samples; If a preset termination event occurs for the first time before the preset observation deadline, the candidate sample is marked with the event occurrence sample identifier, and the service survival time of the candidate sample from the service survival start point to the event occurrence is recorded. Otherwise, the candidate sample is marked with the event non-occurrence sample identifier, and the service survival time of the candidate sample from the service survival start point to the observation deadline is recorded.

[0124] The service duration determination device provided in this application embodiment can execute the method shown in the above method embodiment. Its implementation principle and beneficial effects can be referred to the relevant description in the method embodiment, and will not be repeated here.

[0125] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. For example... Figure 4 As shown, the electronic device includes a memory 401 and at least one processor 402.

[0126] The memory 401 is used to store computer program code, which includes computer instructions. These computer instructions run in the aforementioned electronic device to implement the method shown in the above-described method embodiments. For example, the memory may include high-speed random access memory (RAM), and may also include non-volatile memory (NVM), such as at least one disk storage device, and may also be a USB flash drive, portable hard drive, read-only memory, disk, or optical disc, etc.

[0127] Processor 402 can be a general-purpose processor, including a Central Processing Unit (CPU), a network processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. Processor 402 can also be other general-purpose processors. The general-purpose processor can be a microprocessor or any conventional processor.

[0128] The memory 401 and processor 402 are communicatively connected. For example, the memory 401 can be connected to the processor 402 via a system bus and communicate with it. The system bus can be a peripheral component interconnect (PCI) bus, an extended industry standard architecture (EISA) bus, an industry standard architecture (ISA) bus, etc. The system bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the figure, but this does not mean that there is only one bus or one type of bus.

[0129] Optionally, the memory 401 can be either standalone or integrated with the processor 402. When the memory 401 is set up independently, it is connected to the processor 402 via a system bus.

[0130] This application also provides a chip for executing instructions, which is used to execute the technical solution of the service duration determination method in the above embodiments.

[0131] This application also provides a computer-readable storage medium storing computer instructions. When these computer instructions are executed by a processor, they are used to implement the technical solution of the service duration determination method described in the above embodiments. Specifically, when the computer instructions are executed by a processor, the electronic device can execute the technical solution of the service duration determination method described in the above embodiments.

[0132] This application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When the at least one processor executes the computer program, it can implement the technical solution of the service duration determination method in the above embodiments.

[0133] The aforementioned computer-readable storage media can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The computer-readable storage media can be any available medium accessible to a general-purpose or special-purpose computer.

[0134] An exemplary computer-readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Of course, the computer-readable storage medium can also be a component of the processor. The processor and the computer-readable storage medium can reside in an application-specific integrated circuit (ASIC). Alternatively, the processor and the computer-readable storage medium can exist as discrete components in an electronic control unit or main control device; this application does not limit this.

[0135] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules 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 indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0136] The modules described as separate components may or may not be physically separate. The components shown as modules 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 modules can be selected to implement the solution of this embodiment according to actual needs.

[0137] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit composed of the above modules can be implemented in hardware or in the form of hardware plus software functional units.

[0138] The integrated modules described above, implemented as software functional modules, can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods of the various embodiments of this application.

[0139] It should be understood that the steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor.

[0140] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for determining service duration, characterized in that, The method includes: Obtain basic information about the target object; wherein, the basic information includes the current service duration for which the target object has received resource services, the industry information to which the target object belongs, and the resource information of the target object; Based on the industry information and the resource information, a target model suitable for the target object is determined from a set of preset models; the set of preset models are pre-trained from service data of multiple historical service objects. Obtain resource status feature data of the target object; based on the target model and the resource status feature data, predict the total service life of the target object with the service start point of the resource service as the service life start point, and obtain the predicted total service life; wherein, the total service life represents the duration during which no preset termination event occurs with the service start point as the start time point; The difference between the predicted total service duration and the current service duration is used as the first suggested service duration for the renewal service of the resource service provided to the target object. Based on the pre-assessed service rating information of the target object, the first suggested service duration is calibrated to obtain the target service duration.

2. The service duration determination method according to claim 1, characterized in that, The step of predicting the total service duration of the target object, starting from the service start point of the resource service, based on the target model and the resource status feature data, to obtain the predicted total service duration includes: According to the first expression, determine the first function corresponding to the target object; the first expression is: H(t) = h0(t) × exp(b1×x1+b2×x2+…+bk×xk); Wherein, H(t) is the first function, which is used to characterize the instantaneous probability of the target object experiencing a preset termination event at time t; h0(t) is the benchmark function corresponding to the target model that changes with time t; b1, b2...bk are the weight coefficients in the target model; and x1, x2...xk are the resource status feature data. According to the second expression, the service survival curve corresponding to the target object is obtained; the second expression is: S(t) = exp(-∫H(t) dt); Wherein, S(t) characterizes the service survival curve, and S(t) represents the service survival probability of the target object at time point t, calculated from the service start point of the resource service; Determine the target time point in the service survival curve that corresponds to the preset survival probability, and use the interval between the target time point and the service start point as the predicted total service duration.

3. The service duration determination method according to claim 1, characterized in that, The target model corresponds to a confidence prediction interval, which is defined by an upper limit and a lower limit. The step of calibrating the first suggested service duration based on the pre-assessed service rating information of the target object to obtain the target service duration includes: Based on the service rating information, a second suggested service duration and a service level corresponding to the target object are determined; the service level is used to characterize the probability of a preset termination event occurring. According to the third expression, the maximum recommended service duration for providing the renewal service to the target object is determined; the third expression is: Ta = [Tr / C] × C; Where Ta represents the maximum suggested service duration, Tr represents the first suggested service duration, C is a preset value, and [ ] indicates rounding down; and the preset maximum service duration and preset minimum service duration corresponding to the resource service are obtained; If the predicted total service duration is within the confidence prediction interval, the maximum suggested service duration is adjusted according to the relationship between the first suggested service duration and the second suggested service duration, as well as the service level, to obtain the target service duration. If the predicted total service duration is greater than the upper limit, and the first suggested service duration is greater than the second suggested service duration, then the minimum value between the first suggested service duration and the preset maximum service duration shall be taken as the target service duration. If the predicted total service duration is less than the lower limit, and the first suggested service duration is less than or equal to the second suggested service duration, then the target service duration is determined from the preset minimum service duration and the first suggested service duration according to the service level.

4. The service duration determination method according to claim 3, characterized in that, The service levels include a first level, a second level, a third level, and a fourth level; the probability of the preset termination event occurring increases sequentially from the first level to the second level, the third level, and the fourth level. The step of adjusting the maximum suggested service duration based on the relationship between the first suggested service duration and the second suggested service duration, and the service level, to obtain the target service duration includes: If the first suggested service duration is less than or equal to the second suggested service duration, and the service level corresponding to the target object is the third level, then the difference between the maximum suggested service duration and the preset duration shall be taken as the target service duration. If the first suggested service duration is less than or equal to the second suggested service duration, and the service level corresponding to the target object is the first level, then the sum of the maximum suggested service duration and the preset duration shall be used as the target service duration. If the duration of the first suggested service is greater than the duration of the second suggested service, and the service level is the fourth level, then the target service duration is determined to be zero. If the first suggested service duration is greater than the second suggested service duration, and the service level corresponding to the target object is any one of the first level, the second level, or the third level; or if the first suggested service duration is less than or equal to the second suggested service duration, and the service level corresponding to the target object is the second level, then the maximum suggested service duration shall be taken as the target service duration.

5. The service duration determination method according to claim 4, characterized in that, The step of determining the target service duration based on the service level, between the preset minimum service duration and the first suggested service duration, includes: If the service level corresponding to the target object is the third level, then the minimum value between the first suggested service duration and the preset minimum service duration shall be used as the target service duration; If the service level corresponding to the target object is the first level, then the maximum value between the first suggested service duration and the preset minimum service duration shall be used as the target service duration; If the service level corresponding to the target object is the second level, then the preset minimum service duration is taken as the target service duration.

6. The service duration determination method according to claim 3, characterized in that, The method further includes: Based on service data from multiple historical service objects, the Breuer score of the model at each prediction time point is calculated; the magnitude of the Breuer score is negatively correlated with the prediction accuracy of the model. The continuous time interval in which the Breuer score remains below a preset threshold is used as the confidence prediction interval for the model.

7. The service duration determination method according to claim 2, characterized in that, After obtaining the target service duration, the method further includes: Based on the weight coefficients in the target model, at least one key feature is determined from the resource status feature data; If the weight coefficient corresponding to the key feature is positive, then output the suggestion information to reduce the value corresponding to the key feature; If the weight coefficient corresponding to the key feature is negative, then a suggestion to increase the value of the key feature will be output.

8. The method for determining service duration according to any one of claims 1-7, characterized in that, Before determining the target model suitable for the target object from a predefined set of models based on the industry information and the resource information, the method further includes: Acquire service data for multiple historical service objects; wherein, the service data includes the service start time, service end time, identifier of whether a preset termination event has occurred, industry category information, resource information, and status characteristic data of the service object; Based on the service data of the aforementioned multiple historical service objects, a training sample set and a test sample set are constructed. Based on industry category information and resource information, the training sample set is divided into multiple training subsets and corresponding multiple test subsets; For each training subset and its corresponding test subset, perform the following steps: The historical state feature data of historical service objects in the training subset are used as training input data, and the corresponding service life and whether a preset termination event occurs are used as training labels. Based on the training input data and training labels, a candidate model is trained and obtained; The candidate model is tested using a test subset corresponding to the training subset. If the test results meet the preset performance criteria, the candidate model is determined to be a trained model.

9. The service duration determination method according to claim 8, characterized in that, The historical status feature data includes historical resource attribute feature data and historical resource change feature data; The construction of a training sample set based on the service data of the multiple historical service objects includes: From a pool of historical service recipients, select the service recipients who first request the renewal of the resource service within a preset first time period as candidate samples; For each candidate sample, perform the following processing to obtain the training samples in the training sample set: The service life start point is the time when the corresponding historical service object first applies for resource services, and the observation time point is the time when the historical service object first renews its service. Extract the historical resource attribute feature data of the historical service object at the observation time point, and extract the historical resource change feature data of the historical service object during the observation period; wherein, the observation period is a preset time period from the observation time point forward; The historical resource attribute feature data and the historical resource change feature data are used as the historical state feature data of the candidate samples; If a preset termination event occurs for the first time before the preset observation deadline, the candidate sample is marked with an event occurrence sample identifier, and the service survival time of the candidate sample from the service survival start point to the event occurrence is recorded. Otherwise, the candidate sample is marked with an event non-occurrence sample identifier, and the service survival time of the candidate sample from the service survival start point to the observation deadline is recorded.

10. A service duration determination device, characterized in that, The device includes: An acquisition unit is used to acquire basic information about a target object; wherein, the basic information includes the current service duration for which the target object has received resource services, the industry information to which the target object belongs, and the resource information of the target object; The first processing unit is used to determine a target model suitable for the target object from a set of preset models based on the industry information and the resource information; the set of preset models are pre-trained from service data of multiple historical service objects. The second processing unit is used to acquire resource status feature data of the target object, and based on the target model and the resource status feature data, to predict the total service life of the target object with the service start point of the resource service as the service life start point, and to obtain the predicted total service life; wherein, the total service life represents the duration during which no preset termination event has occurred with the service start point as the start time point. The determining unit is configured to use the difference between the predicted total service duration and the current service duration as the first suggested service duration for providing the resource service renewal service to the target object; The calibration unit is used to calibrate the first suggested service duration based on the service score information of the target object that has been pre-evaluated, so as to obtain the target service duration.

11. An electronic device, characterized in that, include: The electronic device includes a memory and at least one processor; the memory is communicatively connected to the processor; the memory is used to store computer program code, the computer program code including computer instructions; when the processor executes the computer instructions, the electronic device performs the service duration determination method as described in any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, are used to implement the service duration determination method as described in any one of claims 1-9.

13. A computer program product, characterized in that, When the computer program product is run on a computer / executed by the computer's processor, the service duration determination method as described in any one of claims 1-9 is implemented.

Images