Resource processing method and device, equipment and storage medium

Abstract

The invention provides a resource processing method and device, equipment and a storage medium, and relates to the technical field of data processing. According to the scheme, the method comprises the steps of obtaining a multi-dimensional feature attribute and a basic resource allocation quota of a resource requester; constructing a target attribute value combination based on the value of each feature attribute in the multi-dimensional feature attributes, and determining a group type of a target group to which the resource requester belongs; determining a target adjustment coefficient corresponding to the resource requester based on the group type of the target group and a mapping relationship between the group type and the adjustment coefficient; and adjusting the basic resource allocation quota based on the target adjustment coefficient to obtain a target resource allocation quota of the resource requester, and performing resource processing according to the target resource allocation quota. The group type of the resource requester is identified based on the multi-dimensional feature attribute, and the resource quota is adjusted by using the adjustment coefficient obtained through security constraint analysis, so that the flexibility of resource quota decision in a diversity scene is improved, and the balance of security control and quota adaptation is realized.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of data processing, and in particular, to a resource processing method and device, equipment and a storage medium. BACKGROUND

[0002] Resource processing refers to a processing process in which a resource holder performs resource allocation, scheduling or authorization operations according to a resource request of a resource requester.

[0003] Among them, the decision of the resource quota is a key link of the resource processing. At present, the decision method of the resource quota is usually constructed based on a rule engine or a scorecard, a comprehensive score is obtained by calculating the feature information of the resource requester, and the comprehensive score is mapped to the final resource quota according to a mapping rule.

[0004] However, the comprehensive score in the above method is single-dimensional, and lacks flexibility when facing diverse business scenarios, and it is difficult to adapt the quota to different business scenarios according to the security constraints, which affects the balance between the security control and the quota adaptation of the resource quota decision. SUMMARY

[0005] Embodiments of the present application provide a resource processing method, device, equipment and storage medium, by identifying the group type (business scenario) of different resource requesters based on multi-dimensional feature attributes, and adjusting the resource quota of the resource requester based on the adjustment coefficient obtained by analyzing the quantitative relationship between the adjustment coefficient and the bad rate in each group type under the condition of meeting the preset security constraint condition, the problem of insufficient flexibility in dealing with diverse business scenarios in related technologies can be overcome, and the balance between the security control and the quota adaptation of the resource quota decision in different business scenarios can be realized.

[0006] To achieve the above object, the embodiments of the present application adopt the following technical solutions: In a first aspect, a resource processing method is provided, the method comprising: Firstly, multi-dimensional characteristic attributes of the resource requester are acquired, and a basic resource allocation quota is acquired, wherein the multi-dimensional characteristic attributes include a security verification level of the resource requester, a type of the requested resource, a health degree of an application field of the requested resource, and a behavior characteristic of the resource request. Secondly, a target attribute value combination is constructed based on values of the characteristic attributes in the multi-dimensional characteristic attributes, and a group type of a target group to which the resource requester belongs is determined from a preset attribute value combination and group type mapping table. The attribute value combination and the group type are in one-to-one correspondence. Then, a target adjustment coefficient corresponding to the resource requester is determined based on the group type of the target group and a mapping relationship between the group type and the adjustment coefficient. The group type of the target group and the target adjustment coefficient are in one-to-one correspondence. The adjustment coefficient is obtained by constraint analysis on a fitting relationship between the adjustment coefficient and a resource request failure rate. The target adjustment coefficient is used to increase a value of the basic resource allocation quota under a condition that a preset group security condition is met. The resource request failure rate is obtained by counting a number of resource requesters in a corresponding group in a resource return overdue state. Finally, the basic resource allocation quota is adjusted based on the target adjustment coefficient to obtain a target resource allocation quota of the resource requester, and the resource is processed based on the target resource allocation quota.

[0007] The resource processing method provided by the embodiment of the application determines the group type to which the resource requester belongs by constructing the attribute value combination based on the multi-dimensional characteristic attributes of the resource requester. Compared with the method of fusing the multi-dimensional characteristics into a single comprehensive score to make a quota decision, the method can accurately identify the specific business scenario (i.e., the group) in which the resource requester is located by retaining and using the original semantic information of each dimension characteristic and the combination relationship thereof. This fine-grained scenario identification provides accurate classification basis for implementing a differentiated resource allocation strategy. Further, the corresponding target adjustment coefficient is determined from a preset mapping relationship based on the group type. The mapping relationship is established based on the fitting relationship between the adjustment coefficient and the resource request failure rate after constraint analysis. This enables different groups to obtain a differentiated adjustment strategy that matches the security characteristics of the groups and is calibrated by the quantized security result (i.e., the resource request failure rate), thereby overcoming the defect of insufficient flexibility of a single strategy in related methods. Finally, the target adjustment coefficient corresponding to the group to which the resource requester belongs is applied to the basic resource allocation quota to adjust the basic resource allocation quota, thereby realizing fine-grained differentiated quota adaptation for different business scenarios under the security constraint, and effectively maintaining the balance between the security control and the quota adaptation of the resource allocation in the resource quota decision under different business scenarios.

[0008] In a possible implementation of the first aspect, the mapping relationship between the group type and the adjustment coefficient is obtained in the following manner. First, historical resource allocation data of different groups is obtained. The historical resource allocation data includes resource allocation records of different historical resource requesters, and each resource allocation record includes a historical resource allocation amount and a resource return overdue state. Second, the historical resource allocation data corresponding to each group is subjected to feature fitting processing to obtain a quantitative relationship between the adjustment coefficient and the bad rate of each group. Third, based on the quantitative relationship between the adjustment coefficient and the bad rate of each group, the adjustment coefficient corresponding to each group is determined under the constraint condition that an expected resource request bad rate is less than or equal to a preset bad rate threshold of the corresponding group. The expected resource request bad rate is obtained based on the to-be-determined adjustment coefficient through the quantitative relationship. Finally, based on the group type of each group and the adjustment coefficient corresponding to each group, the mapping relationship between the group type and the adjustment coefficient is constructed.

[0009] It should be understood that this implementation objectively fits the quantitative relationship between the adjustment coefficient and the resource allocation security based on historical data, and converts the quota adjustment problem into an optimization solving process under the bad rate constraint. The value of the adjustment coefficient is determined based on verifiable data analysis and mathematical optimization, thereby achieving the scientificity, explainability, and controllability of the security boundary of the quota adjustment strategy.

[0010] In another possible implementation of the first aspect, the feature fitting includes logistic regression fitting. The feature fitting processing of the historical resource allocation data corresponding to each group to obtain the quantitative relationship between the adjustment coefficient and the bad rate of each group includes the following. For any first group in different groups, based on the historical resource allocation data corresponding to the first group, the historical resource allocation amount is taken as a feature, and the resource return overdue state is taken as a label, to perform logistic regression fitting to obtain a logistic regression function. The logistic regression function represents the quantitative relationship between the historical resource allocation amount and the resource request bad rate. The logistic regression function includes a feature weight coefficient and an intercept term. The feature weight coefficient represents the influence degree of the resource allocation amount in the first group on the resource request bad rate. The intercept term represents the influence degree of a non-resource allocation amount factor in the first group on the resource request bad rate. The average resource allocation amount of the first group is determined. The adjustment coefficient corresponding to the first group is obtained by taking the amount obtained by adjusting the average resource allocation amount based on the to-be-determined adjustment coefficient as an independent variable of the logistic regression function.

[0011] It should be understood that this implementation transforms the relationship between resource allocation and non-performing rate into an explicit mathematical function through a logistic regression model. It quantifies the influence of core variables on other factors using feature weight coefficients and intercept terms, while introducing the average resource allocation as a benchmark for the adjustment coefficient. This approach ensures the interpretability of the quantitative relationship and provides a clear functional framework for solving the adjustment coefficient, avoiding the subjectivity of empirical coefficient values ​​and improving the adaptability of the adjustment strategy to actual data.

[0012] In another possible implementation of the first aspect, the preset group security condition includes: the expected resource request failure rate corresponding to the target group is less than or equal to the preset failure rate threshold of the target group.

[0013] It should be understood that this implementation method clarifies the security boundaries of the group (business scenario), and links the expected failure rate with the preset threshold as a security judgment standard. This ensures that resource allocation is always within a safe and controllable range, and provides a pre-constraint for the optimization of adjustment coefficients. It takes into account both individual and global security control, and ensures the stable operation of the resource pool.

[0014] In another possible implementation of the first aspect, based on the quantitative relationship between the adjustment coefficient and the defect rate of each group, and with the constraint that the expected resource request defect rate is less than or equal to the preset defect rate threshold of the corresponding group, the adjustment coefficient corresponding to each group is determined. This includes: for any second group among different groups, with the constraint that the expected resource request defect rate is less than or equal to the preset defect rate threshold of the corresponding group, and with the objective of maximizing the resource allocation amount, constructing the Lagrangian function corresponding to the second group. The adjustment coefficient of the second group is determined based on the numerical optimization algorithm and the Lagrangian function.

[0015] It should be understood that this implementation transforms the solution for the adjustment coefficient into a constrained mathematical optimization problem. A Lagrangian function is constructed with the defect rate not exceeding a threshold as a safety constraint and maximizing resource allocation as the business objective. Numerical optimization algorithms are then used to solve for the optimal adjustment coefficient. This allows the adjustment coefficient to be chosen in a way that satisfies both safety control requirements and maximizes business growth through resource allocation.

[0016] In another possible implementation of the first aspect, the method further includes: for any third group among different groups, obtaining resource allocation data for resource allocation based on the mapping relationship between the application group type and the adjustment coefficient of the third group; determining the actual resource request failure rate of the third group based on the resource allocation data; determining the expected request failure rate of the third group based on the quantitative relationship between the adjustment coefficient and the failure rate of the third group and the adjustment coefficient of the third group; and redetermining the adjustment coefficient of the third group based on the resource allocation data if the deviation rate between the actual resource request failure rate and the expected request failure rate of the third group is greater than or equal to a preset deviation rate threshold.

[0017] It should be understood that this implementation method determines the effectiveness of existing adjustment coefficients by comparing the deviation between the actual and expected defect rates. When the deviation exceeds a threshold, the adjustment coefficients are refitted and determined based on the latest resource allocation data. This ensures that the mapping relationship between group types and adjustment coefficients can adapt to dynamic changes in data distribution, effectively improving the timeliness and adaptability of the adjustment strategy. This ensures that a balance between resource allocation security and business growth can be maintained even when market conditions or user characteristics change.

[0018] In another possible implementation of the first aspect, after determining the actual resource request failure rate of the third group based on resource allocation data, the method further includes: if the ratio of the actual resource request failure rate of the third group to the preset failure rate threshold corresponding to the third group is greater than or equal to a preset warning threshold, updating the adjustment coefficient of the third group in the mapping relationship between group type and adjustment coefficient based on a preset scaling factor. The warning threshold is greater than 1. The preset scaling factor is less than 1.

[0019] It should be understood that this implementation establishes a safety early warning mechanism by setting the ratio of the actual defect rate to a preset threshold as the early warning trigger condition. This mechanism can achieve rapid and conservative adjustments to the resource allocation strategy for this group without relying on complex model reconstruction, so as to suppress the further development of poor performance and thus maintain the safety and controllability of resource allocation.

[0020] Secondly, a resource processing apparatus is provided, the apparatus comprising: The acquisition module is used to acquire the multi-dimensional characteristic attributes of the resource requester and the basic resource allocation quota. The multi-dimensional characteristic attributes include: the security verification level of the resource requester, the type of the requested resource, the health of the domain in which the requested resource is applied, and the behavioral characteristics of the resource request. The processing module is used to construct target attribute value combinations based on the values ​​of each feature attribute in the multi-dimensional feature attributes. It determines the group type of the target group to which the resource requester belongs from a preset attribute value combination and group type mapping table; where each attribute value combination corresponds one-to-one with a group type. Based on the group type of the target group and the mapping relationship between the group type and the adjustment coefficient, it determines the target adjustment coefficient corresponding to the resource requester; the group type of the target group corresponds one-to-one with the target adjustment coefficient; the adjustment coefficient is obtained through constraint analysis of the fitting relationship between the adjustment coefficient and the resource request failure rate in the group; the target adjustment coefficient is used to increase the basic resource allocation quota value while meeting preset group security conditions; the resource request failure rate is obtained by the number of resource requesters in the corresponding group whose resource return overdue status is overdue; the basic resource allocation quota is adjusted based on the target adjustment coefficient to obtain the target resource allocation quota for the resource requester, and resource processing is performed using the target resource allocation quota.

[0021] Thirdly, a resource processing apparatus 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 resource processing apparatus to perform the method as described in the first aspect and any possible implementation thereof.

[0022] Fourthly, a computer-readable storage medium is provided that stores computer instructions. When executed by a processor, the computer instructions are used to implement the method as described in the first aspect and any possible implementation thereof.

[0023] Fifthly, a computer program product is provided that, when run on a computer or executed by a processor of the computer, implements the method described in the first aspect and any possible design thereof. The computer may be the resource processing device described in the third aspect and any possible implementation thereof.

[0024] It is understood that the beneficial effects achieved by the resource processing apparatus described in the second aspect, the resource processing 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, which will not be repeated here. Attached Figure Description

[0025] Figure 1 A flowchart illustrating a resource processing method provided in an embodiment of this application; Figure 2A flowchart illustrating a method for obtaining the mapping relationship between group type and adjustment coefficient provided in an embodiment of this application; Figure 3 A flowchart illustrating a method for obtaining the quantitative relationship between adjustment coefficient and defect rate, provided in an embodiment of this application; Figure 4 A flowchart illustrating a method for determining an adjustment coefficient provided in an embodiment of this application; Figure 5 A flowchart illustrating an adjustment coefficient update method provided in an embodiment of this application; Figure 6 This is a schematic diagram of the structure of a resource processing device provided in an embodiment of this application; Figure 7 This is a schematic diagram of the structure of a resource processing device provided in an embodiment of this application. Detailed Implementation

[0026] 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.

[0027] 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.

[0028] 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.

[0029] 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, they do not mean that the applicant has used or necessarily used the solution.

[0030] Related resource allocation decision-making methods are typically based on compressing features into a single comprehensive score, and then mapping it to the final allocation through a rule engine or scoring card. For example, by predicting the security performance of resource requests, a security outcome score is determined, and then the final allocation is determined based on the security score.

[0031] However, this single-dimensional decision-making approach has significant shortcomings when facing diverse business scenarios: First, because different business scenarios have varying requirements for security constraints and credit limit adaptation, a single scoring method cannot flexibly reflect these differences. Scoring models are often trained based on the overall sample, and their output cannot be finely adjusted for specific scenarios. This leads to being too conservative in some scenarios, which is not conducive to business growth, while being too aggressive in other scenarios, which reduces the security of resource allocation.

[0032] Secondly, this method lacks a dynamic adaptation mechanism for scenario-based security boundaries. Credit limit decisions are essentially a balance between security control and business adaptation, but a single scoring system cannot embed differentiated risk tolerance. In other words, it cannot make targeted and controlled adjustments to the credit limit under specific scenario security constraints (such as a preset upper limit for the non-performing rate), thus affecting the balance and refinement of the overall decision-making.

[0033] Therefore, how to achieve a balance between credit limit adaptation and security control in different scenarios is an urgent problem to be solved.

[0034] In view of this, this application provides a resource processing method that, based on the original resource quota decision method, introduces a business scenario identification mechanism based on multi-dimensional feature combination and a differentiated quota adjustment mechanism based on scenario-based security constraints (characterized by quantified failure rate) to achieve a balance between quota adaptation and security control under different business scenarios.

[0035] The resource processing method provided in this application can be applied to computing devices. Specifically, the computing device can be a single server or a server cluster composed of multiple servers, or a computer, or a processor or processing chip in a server or computer, etc. This application does not limit the specific device form of the computing device.

[0036] like Figure 1 As shown, the resource processing method provided in this application embodiment, when applied to the above-mentioned computing device, specifically includes the following steps S101-S104: S101. Obtain the multi-dimensional characteristic attributes of the resource requester and the basic resource allocation quota.

[0037] Among them, the resource requester refers to the entity that initiates the request to acquire resources, which is usually an identified individual user, enterprise or organization.

[0038] Multidimensional feature attributes refer to a structured dataset that describes a resource requester and its current request from multiple perspectives. The various dimensions of this dataset work together to provide comprehensive information input for subsequent scene recognition and differentiated decision-making. Multidimensional feature attributes include: the resource requester's security verification level, the type of resource requested, the health of the domain in which the requested resource is applied, and the behavioral characteristics of the resource request.

[0039] The security verification level of a resource requester reflects the authenticity of its identity and its historical credit security level. A higher level generally indicates more stringent security checks and a more reliable historical performance record. The type of requested resource reflects the nature, purpose, and potential value of the resource itself; different types of resources correspond to different security requirements and benefit characteristics. The health of the application field of the requested resource reflects the overall security status and development stability of the industry or scenario where the resource is intended to be used. The behavioral characteristics of resource requests reflect the requester's historical resource requests and usage patterns, used to characterize its behavioral habits and potential risk propensities.

[0040] The basic resource allocation quota refers to the initial resource quota obtained based on the access approval process for resource requesters. It usually serves as a benchmark value, providing a starting point for subsequent scenario-based and refined adjustments.

[0041] One possible implementation is that the computing device can obtain the corresponding data values ​​of the aforementioned multidimensional feature attributes by calling the internal business database, the interface of the external regulatory agency, and the business application form.

[0042] One possible implementation is that the computing device can process the resource requester's resource request through a scoring model or rule engine in a relevant method, and output a basic resource allocation quota that conforms to existing business logic and regulatory compliance. Then, the computing device performs further scenario-specific quota adaptation adjustments based on this basic resource allocation quota.

[0043] Specifically, adopting entirely new machine learning models often requires replacing existing systems, resulting in high implementation costs and regulatory challenges due to insufficient model interpretability. While scorecard models offer strong interpretability, their rules are rigid and difficult to adapt quickly to market changes. Therefore, this approach can reuse relevant credit limit decision-making frameworks, reducing the complexity of modifications, while enhancing the balance between security control and resource adaptation in credit limit decisions by overlaying multi-dimensional scenario recognition and adjustment coefficients.

[0044] For example, in the financial sector, the requested resource could be a credit loan, with corresponding multi-dimensional data including: security verification level (e.g., customer risk level, such as R3); resource type (unsecured revolving loan); application field health (the prosperity index of the industry to which the funds are invested); and behavioral characteristics (the customer's recent credit card usage and repayment history).

[0045] For example, in the internet field, requested resources could be cloud computing GPU quotas, with corresponding multi-dimensional data including: security verification level (account multi-factor authentication status); resource type (A100 computing instance); application domain health (activity level of the technical field of the AI ​​tasks being undertaken); and behavioral characteristics (historical load curve and expansion frequency of the account).

[0046] For example, in the automotive sector, the requested resource could be a short-term rental of a new energy vehicle, with corresponding multi-dimensional data including: safety verification level (driver's license verification results and platform credit score); resource type (long-range version of the vehicle); application domain health (traffic and charging infrastructure status in the rental city); and behavioral characteristics (driving time periods and speed fluctuations in the user's historical orders).

[0047] This application does not limit the specific type of requested resources, and correspondingly, it does not limit the application scenarios.

[0048] S102. Construct a target attribute value combination based on the values ​​of each feature attribute in the multidimensional feature attributes, and determine the group type of the target group to which the resource requester belongs from the preset attribute value combination and group type mapping table.

[0049] Among them, the target attribute value combination refers to the vector or data tuple formed by the resource requester's values ​​on various dimensional characteristic attributes in a predetermined order.

[0050] A group refers to a collection of resource requesters with the same or similar combinations of characteristic attribute values. Since the combination of attribute values ​​includes the security status of the resource requester, the type of resource requested, the purpose of the resource, and the behavior pattern, each group essentially corresponds to a business scenario pattern (the group type is also the label of the business scenario pattern), that is, the mapping relationship between attribute value combinations and group types.

[0051] In some embodiments, the feature values ​​of multidimensional feature attributes are obtained by integrating and parsing multi-source information. The computing device can either reuse existing credit decision-making methods (such as its scoring model or rule engine) to process the resource requester's basic information, or it can directly process the resource requester's basic information to obtain the attribute values ​​of the multidimensional feature attributes.

[0052] One possible implementation is that, for security verification levels, the computing device can directly extract them from the relevant scoring model or rule engine.

[0053] For example, in the financial field, assuming that the score range for the security verification of the resource requester is [0, 1000], the computing device divides it into five discrete level intervals: AAA level (score ≥ 750 points), AA level (700 ≤ score < 750 points), A level (650 ≤ score < 700 points), B level (600 ≤ score < 650 points), and C level (score < 600 points).

[0054] One possible implementation is that, for the type of resource requested, the computing device can directly take the type of resource selected or specified by the resource requester when making the resource request. The resource type itself is a predefined discrete enumeration value.

[0055] For example, in the financial field, the requested resource can be a financial product, and the type of requested resource is also the type of financial product. An exemplary classification is: P1 - consumer loans (including home renovation loans, travel loans, etc.), P2 - business loans (loans to micro and small enterprises, loans to individual businesses, etc.), P3 - credit card overdrafts, P4 - mortgage loans, and P5 - auto loans.

[0056] One possible implementation involves determining the health of the domain to which the requested resource applies, based on the domain type of the application and by obtaining relevant information about that domain. Specifically, the computing device can determine the health level by weighting and integrating current macroeconomic indices, the prosperity index of the application domain, and the quantitative value of the economic level of the region where the resource requester is located.

[0057] For example, a computing device can determine its health status by using a formula: 0.4 × macroeconomic index + 0.3 × business climate index + 0.3 × economic level quantification value, where 0.5, 0.3, and 0.2 are their respective weights.

[0058] One possible implementation is that, based on the behavioral characteristics of resource requests, the computing device can obtain a historical resource return score from the business database based on the recent resource return records of the resource requester, calculate an account activity score based on the usage frequency and resource amount changes of the resource requester's account, and calculate a credit trend score based on the historical changes in fixed resources and the amount of resources to be returned.

[0059] For example, a computing device can determine the behavioral characteristics of a resource request by using the following formula: 0.5 × historical resource return score + 0.3 × account activity score + 0.2 × creditworthiness trend score, where 0.5, 0.3, and 0.2 are their respective weights.

[0060] The embodiments of this application do not limit the specific method for obtaining multidimensional feature attribute values.

[0061] In some embodiments, based on the differences in the data sources of the attribute values ​​of the aforementioned feature attributes (such as security verification level and resource type being predefined discrete enumeration values, while application domain health and behavioral characteristics are continuous values ​​calculated by a model), the computing device can first discretize the continuous values ​​during the process of constructing the attribute value combination, and then combine the discretized feature attribute values.

[0062] One possible implementation is that the computing device can discretize the health status and behavioral characteristics of the application domain based on a preset threshold range and a piecewise function.

[0063] For example, for the health level (e) of the application domain, the computing device presets three threshold ranges: when e≥80, it is discretized into a high level (H); when 60≤e<80, it is discretized into a medium level (M); when e<60, it is discretized into a low level (L).

[0064] Similarly, for the behavioral characteristics (b) of resource requests, the computing device also pre-defines similar rules: when b≥80, it is discretized as excellent; when 60≤b<80, it is discretized as good; when b<60, it is discretized as average.

[0065] For example, the above combination of target attribute values ​​can be represented as M{r,p,e,b}. Here, r represents the security verification level, with a value range of {AAA, AA, A, B, C}; p represents the type of requested resource, with a value range of {P1, P2, P3, P4, P5}; e represents the health of the domain to which the requested resource applies, with a value range that can be mapped to {H, M, L}; and b represents the behavioral characteristics of the resource request, with a value range that can be mapped to {Excellent, Good, Average}.

[0066] In some embodiments, the computing device obtains the mapping table by enumerating the value ranges of each feature attribute. Specifically, the computing device first determines all possible values ​​of each feature attribute (such as security verification level) after discretization, forming a finite value range. Then, it calculates the Cartesian product of these value ranges, enumerating all possible combinations of attribute values. For each combination, the computing device assigns a business-meaning group type label based on preset classification rules or clustering results of security performance and behavior patterns of requesters with the same combination in historical data. Finally, the mapping table is obtained by structurally storing all combinations and their assigned group labels into a queryable data table.

[0067] In some embodiments, the dimensions of the multidimensional feature attributes are expandable. Specifically, the computing device can add new feature dimensions to the dimension set (e.g., add a 'seasonality index of resource requests') or adjust the granularity of the value range division of existing dimensions according to business development needs. By updating the construction logic of the mapping table between the attribute value combinations and group types, the computing device can enable the resource processing method to adaptively cover new business scenarios.

[0068] It should be understood that computing devices determine group types by constructing and matching combinations of target attribute values, preserving the inter-feature correlation information and complete semantics lost when compressing scores or rules in a single dimension. This enables precise identification and differentiation of different business scenarios, providing a fundamental basis for implementing differentiated resource allocation strategies that match the security levels of each scenario, and enhancing flexibility in responding to complex business needs.

[0069] S103. Based on the group type of the target group and the mapping relationship between the group type and the adjustment coefficient, determine the target adjustment coefficient corresponding to the resource requester.

[0070] In this system, there is a one-to-one correspondence between the target group type and the target adjustment coefficient. That is, each group corresponds to a specific adjustment coefficient. By distinguishing the adjustment coefficients for each group type, the computing device can optimize credit limit decisions for different business scenarios.

[0071] The target adjustment coefficient is used to increase the basic resource allocation limit while meeting preset group security conditions. By applying differentiated adjustment coefficients to different groups, computing devices can specifically increase resource allocation limits for each group's business scenario, ensuring that its security level is controlled within preset security boundaries (conditions). This improves the adaptability of resource allocation limits to business scenarios and effectively overcomes the problem in related methods that rely on comprehensive scoring and struggle to balance security control and limit adaptation in diverse scenarios.

[0072] In some embodiments, the preset group security condition can be expressed as: the expected resource request failure rate corresponding to the target group is less than or equal to the preset failure rate threshold of the target group. The expected resource request failure rate is determined based on the quantitative relationship between the target adjustment coefficient and the adjustment coefficient corresponding to the target group and the failure rate.

[0073] It should be understood that this implementation method clarifies the security boundaries of the group (business scenario), and links the expected failure rate with the preset threshold as a security judgment standard. This ensures that resource allocation is always within a safe and controllable range, and provides a pre-constraint for the optimization of adjustment coefficients. It takes into account both individual and global security control, and ensures the stable operation of the resource pool.

[0074] The adjustment coefficient is obtained through constraint analysis of the fitted relationship between the adjustment coefficient and the resource request failure rate within a group. Specifically, the computing device analyzes historical data of the group to fit a quantitative relationship between changes in the adjustment coefficient and the resource request failure rate, which can objectively reflect the trend of the impact of quota adjustments on the failure rate within each group. Furthermore, the computing device performs constraint analysis on the above relationship under the premise of the failure rate constraint to derive the adjustment coefficient.

[0075] The above process ensures that the final determined adjustment coefficient inherently satisfies the dual objectives of safety control (constrained by the non-performing rate threshold) and credit limit increase (solved under constraints), thus possessing the ability to optimize credit limit adaptation under the premise of safety and controllability.

[0076] The resource request failure rate is obtained by counting the number of resource requesters in the corresponding group whose resource return overdue status is overdue. The overdue status refers to the status indicator that the resource requester, after obtaining the resource, failed to complete the resource return or fulfill the agreed behavior within the agreed period.

[0077] One possible implementation is that the computing device identifies overdue resource return statuses marked on each resource request record in the group, determines overdue records as bad records, and then calculates the proportion of bad records in the group to the total number of records as the bad resource request rate of the group.

[0078] For details on obtaining the mapping relationship between group type and adjustment coefficient, please refer to the following text. Figure 2 The details of the matter will not be elaborated here.

[0079] S104. Adjust the basic resource allocation quota based on the target adjustment coefficient to obtain the target resource allocation quota for the resource requester, and process resources based on the target resource allocation quota.

[0080] Resource processing refers to the subsequent operations such as resource allocation, permission granting, or service activation based on the determined target resource allocation amount. This application's embodiments do not limit the specific content or process of resource processing.

[0081] One possible implementation is that the computing device has a built-in quota adjustment engine. This engine receives the basic resource allocation quota and the target adjustment coefficient as input parameters, calls a predefined multiplication calculation module for processing, and outputs the target resource allocation quota.

[0082] For example, suppose a resource requester's basic resource allocation quota is 10,000 units of resources, and the target adjustment coefficient for its group is 1.2. The computing device performs a multiplication operation: 10,000 × 1.2 = 12,000, and the resulting target resource allocation quota is 12,000 units of resources.

[0083] In some embodiments, the computing device may also introduce a quota review logic after obtaining the target resource allocation quota and before triggering the resource processing flow. The computing device may compare the target resource allocation quota with the static quota limit for the resource requester (e.g., the maximum grantable quota based on its qualifications). If the target quota does not exceed the static quota limit, subsequent processing is triggered normally; if it exceeds the static quota limit, the static quota limit is used as the final target resource allocation quota.

[0084] This application provides a resource processing method that constructs attribute value combinations based on the multi-dimensional feature attributes of resource requesters to determine their group type. Compared to methods that fuse multi-dimensional features into a single comprehensive score for quota decisions, this method accurately identifies the specific business scenario (i.e., group) of the resource requester by retaining and utilizing the original semantic information of each dimension's features and their combination relationships. This fine-grained scenario identification provides a precise classification basis for implementing differentiated resource allocation strategies. Furthermore, based on the group type, a corresponding target adjustment coefficient is determined from a preset mapping relationship. This mapping relationship is established after constraint analysis based on the fitting relationship between the adjustment coefficient and the resource request failure rate. This allows different groups to obtain differentiated adjustment strategies that match their security characteristics and are calibrated by quantified security results (i.e., resource request failure rate), overcoming the lack of flexibility of single strategies in related methods. Finally, by applying the target adjustment coefficient corresponding to the group to which the resource requester belongs to its basic resource allocation quota, it is possible to achieve refined and differentiated quota adaptation for security constraints in different business scenarios, effectively maintaining the balance between security control and quota adaptation in resource allocation under different business scenarios.

[0085] In some embodiments, the computing device can determine the quantitative relationship between the adjustment coefficient of each group and the resource request failure rate using historical data of each group, as detailed in the following process: Figure 2 As shown, the specific steps include the following: S201-S204: S201. Obtain historical resource allocation data for each of the different groups.

[0086] The historical resource allocation data includes resource allocation records from different historical resource requesters. These records include historical resource allocation amounts and overdue resource return status.

[0087] Specifically, the historical resource allocation quota refers to the amount of resources allocated to the corresponding resource requester, and the overdue resource return status refers to whether the historical requester has fulfilled its return obligation within the agreed period after obtaining the resources. When the computing device aggregates multiple resource allocation records of a group, it can observe the correlation pattern between different allocation quota levels and different overdue occurrence ratios (i.e., resource request failure rates).

[0088] In some embodiments, the computing device obtains data by initiating a structured query to a business database. The computing device constructs a query statement, specifies the required data fields (group type label, historical resource allocation quota, and resource return overdue status), and sets filtering conditions such as time range to obtain historical resource allocation data from the business database.

[0089] For example, the time range can be 12 months, 24 months or 36 months, etc., and this application embodiment does not limit the historical duration of historical resource allocation data.

[0090] In some embodiments, there may be a situation where the historical resource allocation data corresponding to a certain group is too sparse. To address this issue, the computing device can use a preset similarity measurement method to find similar groups with sufficient historical data for the target group with sparse historical data, and supplement the target group with the historical resource allocation data of the similar groups to construct an enhanced dataset for feature fitting.

[0091] One possible implementation is that the computing device uses weighted Euclidean distance as a similarity metric to determine the comprehensive differences between the target group and the candidate group in dimensions such as security verification level, requested resource type, application domain health, and behavioral characteristics, and then derives a similarity score based on the preset weights of each dimension.

[0092] For example, the process of determining the similarity measure between the target group and the candidate group using weighted Euclidean distance is as follows:

[0093] in, Indicates the target customer group with candidate customer groups Similarity score; These represent the attribute values ​​(which have been discretized) for security verification level, requested resource type, application domain health, and behavioral characteristics, respectively. These are the preset weights for the corresponding dimensions, and satisfy the following conditions: .

[0094] For example, a target group has only 150 historical records due to its unique dimensional combination. The computing device finds that it has a similarity score of 0.85 with another group that has 800 historical records and a similar dimensional combination. The computing device can then supplement the former's dataset with some of the latter's historical data.

[0095] S202. Perform feature fitting processing on the historical resource allocation data corresponding to each group to obtain the quantitative relationship between the adjustment coefficient and the failure rate of each group.

[0096] Feature fitting refers to the process of using mathematical or statistical models to find and establish a functional relationship between input variables (features) and output variables (labels) from historical data. In this step, the input variable is the adjustment factor (represented as the adjustment coefficient) related to the resource allocation quota, and the output variable is the resource request default rate calculated from the overdue status of resource returns. Through feature fitting, the implicit influence of quota adjustments on the default rate in historical experience can be explicitly expressed as a calculable quantitative model.

[0097] In some embodiments, the quantitative relationship between the adjustment coefficient and the defect rate specifically includes the following types: monotonically increasing relationship, monotonically decreasing relationship, and nonlinear complex relationship. This relationship reveals the specific direction and extent of the impact of changes in the adjustment coefficient on the resource request defect rate.

[0098] For example, in routine business scenarios that are sensitive to security, increasing the adjustment coefficient (i.e., increasing the allocation amount) usually leads to an increase in the expected non-performing rate, which is reflected as a positively correlated monotonically increasing relationship.

[0099] For example, in some scenarios where there is a strong synergistic effect or a significant positive feedback mechanism between resource input and output benefits, it may be helpful for the resource requester to make more effective use of resources and improve its business conditions in the early stage. This may reduce its probability of default, thus showing a monotonically decreasing relationship between the adjustment coefficient and the default rate within a specific range.

[0100] In some embodiments, the quantitative relationship between the adjustment coefficient and the defect rate may include a mathematical function or be determined by training a machine learning model. Specifically, this quantitative relationship may also be expressed as a function expression or a trained model object (such as a gradient boosting tree model, neural network model, etc.) that takes the adjustment coefficient as input and outputs the expected defect rate.

[0101] One possible implementation is that the computing device can also use a logistic regression algorithm for feature fitting. For details, please refer to the following text. Figure 3 The details of the matter will not be elaborated here.

[0102] Another possible implementation involves the computing device employing machine learning models such as gradient boosting decision trees for feature fitting. Specifically, the computing device uses historical resource allocation data for each group as a training set. Each training sample uses the historical resource allocation amount as an input feature, and the corresponding overdue resource return status (overdue = 1, not overdue = 0) as a binary label, or the aggregated historical resource request failure rate within the group as a continuous value label. The computing device uses a gradient boosting decision tree algorithm for supervised learning on this training set, with the training objective being to teach the model the mapping relationship between the allocation amount and the group's resource request failure rate. After training, a trained prediction model is obtained. To establish a quantitative relationship between the adjustment coefficient and the failure rate, the computing device multiplies the group's baseline allocation amount (e.g., average historical allocation amount) with the adjustment coefficient to be tested, and inputs the product as a new allocation feature into the prediction model. The model's output prediction value is the expected resource request failure rate under that adjustment coefficient. By iterating through or calculating the prediction values ​​under different adjustment coefficients, the quantitative relationship between the two can be obtained.

[0103] S203. Based on the quantitative relationship between the adjustment coefficient and the defect rate of each group, and with the expectation that the defect rate of resource requests is less than or equal to the preset defect rate threshold of the corresponding group as a constraint, determine the adjustment coefficient corresponding to each group.

[0104] The expected resource request failure rate is obtained by mapping a quantitative relationship based on an adjustment coefficient to be determined. The preset failure rate threshold is an acceptable performance level boundary value pre-set for each group, i.e., a safety control parameter, used to ensure the robustness of the resource allocation strategy.

[0105] Specifically, the computing device, for each group, utilizes the quantified relationship between its adjustment coefficient and defect rate, which is expressed as follows: ,in Represents the adjustment coefficient. This represents the expected resource request failure rate under the adjustment coefficient. Additionally, each group has a preset failure rate threshold. Computing devices As a safety constraint, constraint analysis is performed under this constraint to determine the adjustment coefficient corresponding to this group. .

[0106] In some embodiments, the quantitative relationship between the adjustment coefficient and the defect rate is as follows: In the case of a monotonic function, the computing device can directly determine the adjustment coefficient for each group based on the monotonicity of the function.

[0107] One possible implementation, if Follow The equation increases monotonically and can be solved directly by computing devices. Obtain the critical adjustment coefficient Since the function is monotonically increasing, it satisfies the safety constraints. The range of values ​​for the adjustment coefficient is... The computing device can select a value within this range as the adjustment coefficient for the group based on business rules (e.g., to maximize resource allocation). A typical approach is to directly select... .

[0108] For example, taking the (AA, P1, H, Excellent) combination as an example, its historical average resource allocation is... 10,000, target non-performing rate upper limit The defect rate function is based on a logistic regression model: .

[0109] Assuming the fit is obtained This function follows Monotonically increasing. The computing device uses Newton's method to solve the problem. ,from Start iterating until convergence to the optimal solution. .at this time, The credit limit was increased to Ten thousand yuan.

[0110] like Follow If the coefficient increases and then monotonically decreases, increasing the adjustment coefficient will reduce the expected defect rate. In this case, the safety constraint... Typically, the constraints are either automatically satisfied or very weak. The strategy for determining computing devices can focus on business objectives; for example, directly using a preset, high coefficient value (such as a preset maximum coefficient value). Alternatively, under the premise of satisfying other boundary conditions (such as the maximum increase the system can withstand), select the largest possible value. value.

[0111] In other embodiments, the relationship between the adjustment coefficient and the defect rate is quantified. For non-monotonic or complex cases, the computing device can use numerical optimization algorithms to determine the adjustment coefficients for each group.

[0112] Specifically, the problem of optimizing computing device construction: the objective is to maximize the adjustment coefficient. (or maximize the adjusted limit), with the following constraints: and the possible range of coefficient values The computing device uses numerical optimization algorithms such as the Lagrange multiplier method to iteratively solve the problem, ultimately obtaining an adjustment coefficient that satisfies all constraints and makes the objective function optimal (or nearly optimal). The process can be referenced below. Figure 4 The details of the matter will not be elaborated here.

[0113] S204. Based on the group type of each group and the corresponding adjustment coefficient of each group, construct a mapping relationship between group type and adjustment coefficient.

[0114] Among them, building a mapping relationship refers to creating and storing a data structure that can associate each specific group type identifier (i.e., business scenario label) with its optimal adjustment coefficient value obtained after quantitative analysis and security constraint solution.

[0115] For example, this data structure can be represented as a matrix, a hash table, or a relational database table. The matrix form can be a two-dimensional matrix with group type indexes as rows and adjustment coefficients as columns. The hash table form stores data with group type as the key and adjustment coefficient as the value. The relational database table form contains two fields: group type and adjustment coefficient, with each record storing a mapping relationship. This application embodiment does not limit the mapping structure between group type and adjustment coefficient.

[0116] It should be understood that this implementation method objectively fits the quantitative relationship between the adjustment coefficients of each group and the resource request default rate using historical data, and transforms the quota adjustment problem into an optimization solution process under a clear default rate constraint. This ensures that the coefficient values ​​are based on verifiable data analysis and mathematical optimization, thereby improving the scientific nature, interpretability, and controllability of the safety boundary of the adjustment strategy.

[0117] In some embodiments, during the feature fitting process of historical resource allocation data corresponding to each group in S202, the computing device may use logistic regression fitting to obtain the quantitative relationship between the adjustment coefficient and the defect rate of each group. This process is as follows: Figure 3 As shown, S202 specifically includes steps S301-S303: S301. For any first group in different groups, based on the historical resource allocation data corresponding to the first group, with the historical resource allocation amount as the feature and the overdue status of resource return as the label, perform logistic regression fitting to obtain the logistic regression function.

[0118] The logistic regression function represents the quantitative relationship between historical resource allocation limits and resource request failure rates. The logistic regression function includes feature weight coefficients and an intercept term. The feature weight coefficients represent the degree of influence of resource allocation limits on the resource request failure rate within the first group. The intercept term represents the degree of influence of non-resource allocation limit factors on the resource request failure rate within the first group.

[0119] Specifically, the feature weight coefficient can be understood as the strength of the impact of credit limit changes on the logarithm of the default rate; its sign determines the direction of the impact, and its absolute value reflects the sensitivity. The intercept term can be understood as the logarithmic probability of the group's basic security level when the credit limit is zero, reflecting the overall impact of other stable factors besides the credit limit.

[0120] Because historical resource allocation data only contains historical resource quotas and their corresponding performance results, it lacks records that directly reflect the effect of the adjustment coefficient. Furthermore, the adjustment coefficient is a scalar multiplier used to scale the base resource allocation quota value, without changing the business semantics of the quota itself. Therefore, the computing device can first establish a basic relationship between historical resource allocation quotas and non-performing loan rates through logistic regression fitting. Then, by applying the adjustment coefficient to the baseline quota (such as the average quota) in the group, and substituting this variable into the aforementioned basic relationship, a quantitative relationship between the adjustment coefficient and the non-performing loan rate can be derived.

[0121] Specifically, for the historical resource allocation data corresponding to the first group, the process of performing logistic regression fitting with historical resource allocation amount as feature and overdue resource return status as label is as follows: The computing device uses historical resource allocation amount as the only input feature of the model and binary overdue resource return status as the prediction target. Using the maximum likelihood estimation principle, a set of model parameters (i.e. feature weight coefficients and intercept term) is found through iterative optimization algorithm so that the likelihood of the overdue probability distribution predicted by the model is maximized with the historical actual status distribution, thereby completing the model training and obtaining the logistic regression function.

[0122] In some embodiments, the iterative optimization algorithm described above can be gradient descent to obtain the feature weight coefficients and intercept term. After the computing device initializes the model parameters (i.e., the initial values ​​of the feature weight coefficients and intercept term), it calculates the gradient of the prediction error (loss function) of the entire training set or a batch of samples with respect to the model parameters in each iteration, and updates the parameters in the opposite direction of the gradient until the loss function converges to the minimum value or reaches the preset number of iterations. The parameters obtained at this time are the final feature weight coefficients and intercept term.

[0123] For example, if the learning rate is set to 0.01 and the gradient is calculated using all the data in the first group, after 5000 iterations, the loss function no longer decreases significantly. At this point, the feature weight coefficient of 0.15 and the intercept term of -3.2 are the fitting results of the logistic regression function.

[0124] In other embodiments, to solve for the feature weight coefficients and intercept terms more efficiently and accurately, the above iterative optimization algorithm can also be a quasi-Newton method. During the iteration process, the computing device not only utilizes gradient information but also estimates the curvature of the loss function by approximating the Hessian matrix, thereby enabling faster convergence to the optimal parameter solution and obtaining the final feature weight coefficients and intercept terms. This is particularly suitable for scenarios with low feature dimensionality but requiring high-precision fitting.

[0125] For example, computing devices using the L-BFGS algorithm (a quasi-Newton method) can typically converge with higher accuracy in less than 1000 iterations, obtaining feature weight coefficients of 0.152 and intercept terms of -3.21.

[0126] S302. Determine the average resource allocation amount for the first group.

[0127] The average resource allocation amount refers to a measure of the central tendency of the allocation amount (usually an arithmetic mean) calculated based on the historical resource allocation data of the first group. It serves as the baseline value for the allocation amount of this group, used to associate an abstract adjustment coefficient with a specific allocation size.

[0128] Specifically, after obtaining the basic relationship between credit limit and non-performing loan rate, the calculation device needs to determine a benchmark for credit limit adjustment. The average credit limit reflects the typical credit limit size of the group. Therefore, by calculating this average value, the calculation device provides a stable and business-representative input for substituting the adjustment coefficient into the quantitative relationship, ensuring that the derived "adjustment coefficient-non-performing loan rate" relationship is consistent with the actual credit limit level.

[0129] One possible implementation is that the computing device uses an arithmetic mean method for calculation. The computing device extracts the "historical resource allocation quota" field from all historical resource allocation records in the first group, sums these values, divides them by the total number of records, and obtains the arithmetic mean as the average resource allocation quota.

[0130] For example, assuming the first group has 1,000 historical records and the total historical resource allocation is 50,000,000 units of resources, the computing device calculates the average resource allocation as 50,000,000 / 1,000 = 50,000 units of resources.

[0131] In some embodiments, the computing device may also employ a weighted average or a truncated average. For example, for groups with extremely uneven credit limit distributions, the computing device may assign different weights based on the time of recording or the importance of the business to perform a weighted average; or, to eliminate the influence of extreme values, the highest and lowest specific proportions of data may be removed after sorting before calculating the average.

[0132] S303. Using the adjusted amount of the average resource allocation based on the adjustment coefficient to be determined as the independent variable of the logistic regression function, the quantitative relationship between the adjustment coefficient and the non-performing rate corresponding to the first group is obtained.

[0133] Specifically, using the adjusted amount of the average resource allocation based on the adjustment coefficient to be determined as the independent variable of the logistic regression function means substituting the "adjustment coefficient to be determined" and the "average resource allocation amount of the first group" as input variables into the logistic regression function obtained in step S301. This operation realizes the evolution from the "quantitative relationship between resource allocation amount and non-performing rate" to the "quantitative relationship between adjustment coefficient and non-performing rate".

[0134] Specifically, based on step S301, the computing device obtained a quantitative relationship between historical resource allocation quotas and resource request failure rates: ( ), among which, among which, Indicates the historical allocation amount of resources. For the intercept term, These are the feature weight coefficients (i.e., the coefficients of the quota variable). Furthermore, let the average resource allocation amount for this group be... The adjustment coefficient to be determined is The adjustment coefficient is applied to the average amount to obtain the adjusted average resource allocation amount. Computing devices will Substituting into the above quantitative relationship In the middle, we get:

[0135] It should be understood that this implementation transforms the relationship between resource allocation and non-performing rate into an explicit mathematical function through a logistic regression model. It quantifies the influence of core variables on other factors using feature weight coefficients and intercept terms, while introducing the average resource allocation as a benchmark for the adjustment coefficient. This approach ensures the interpretability of the quantitative relationship and provides a clear functional framework for solving the adjustment coefficient, avoiding the subjectivity of empirical coefficient values ​​and improving the adaptability of the adjustment strategy to actual data.

[0136] In some embodiments, regarding S203, the quantitative relationship between the adjustment coefficient and the defect rate D is... When the problem is non-monotonic or complex in form, and requires optimization with the primary constraint that the expected resource request failure rate is less than or equal to a preset failure rate threshold, the process by which the computing device uses the Lagrange multiplier method to construct and solve the constrained optimization problem is as follows: Figure 4 As shown, S203 specifically includes the following steps S401-S402: S401. For any second group in different groups, with the expected resource request failure rate being less than or equal to the preset failure rate threshold of the corresponding group as a constraint, and with the goal of maximizing the resource allocation amount, construct the Lagrangian function corresponding to the second group.

[0137] Specifically, the computing equipment for the second group already possesses a quantitative relationship between its adjustment coefficient and defect rate. and preset defect rate threshold In order to meet safety constraints Under the premise of maximizing resource allocation, the computing device formalizes this business optimization problem as a mathematical optimization problem with inequality constraints. Therefore, the computing device incorporates Lagrange multipliers into the objective function to construct a Lagrange function, providing a standardized mathematical model for subsequent solution using optimization theory.

[0138] One possible implementation is a Lagrangian function constructed by a computing device. The format is as follows:

[0139] in, This represents maximizing the adjusted resource allocation amount (i.e. ) as the target, This serves as a reference value for the basic resource allocation quota of the second group (such as the average resource allocation quota). These are Lagrange multipliers, used to balance the objective function and constraints. At the optimal solution, complementary relaxation conditions must be satisfied. .

[0140] In some embodiments, in addition to the non-performing loan ratio constraint, the computing device may also incorporate other security conditions as constraints, such as quota concentration constraints, maximum single-account exposure constraints, or short-term liquidity stress test indicator constraints.

[0141] Specifically, the quota concentration constraint aims to prevent excessive concentration of resources in a single customer group or a few requesters. This can be expressed as the proportion of the adjusted total expected quota for that group to the total resource pool must not exceed a preset threshold. The maximum single-account exposure constraint limits the upper limit of the quota granted to a single resource requester (or its affiliates), which typically translates into an implicit constraint on the adjustment coefficient α (calculated by influencing the average quota). The short-term liquidity stress test indicator constraint requires that, under simulated specific stress scenarios, the potential rate or scale of resource outflows that may be triggered by the adjusted quota for that group must be below a safety threshold.

[0142] S402. Based on numerical optimization algorithms and the Lagrange function, determine the adjustment coefficients for the second group.

[0143] Specifically, due to the quantitative relationship between the adjustment coefficient of the second group and the defect rate... The problem may exhibit complex forms such as non-monotonicity, making it impossible to obtain the optimal solution by directly solving the equations. Therefore, the computing device, based on a pre-constructed Lagrangian function, uses a numerical optimization algorithm for iterative search. This allows for the systematic determination of (locally) optimal adjustment coefficients that satisfy preset safety conditions, while comprehensively considering the objective function and complex constraints. .

[0144] In some embodiments, in determining the adjustment coefficients based on numerical optimization algorithms and Lagrange functions, the computing device may employ optimization algorithms capable of handling non-convex and nonlinear problems.

[0145] One possible implementation involves the computing device employing a sequential quadratic programming algorithm. The computing device uses the KKT conditions of the Lagrange function as the solution objective. In each iteration, it calculates the first and second derivatives of the objective and constraint functions, constructs and solves a quadratic programming subproblem to update the estimates of the adjustment coefficients and Lagrange multipliers, until the solution converges.

[0146] For example, suppose a group, after fitting historical data, yields an extended logistic regression function containing a nonlinear term, with the defect rate function being: .

[0147] In this function, due to the existence This item, function It is not simply a monotonic function. Analysis shows that... Domain (e.g.) )Inside, Follow The increase exhibits a complex pattern of first rising slowly, then falling, and then rising rapidly again, meaning there exists a "groove" range with a relatively low level of safety.

[0148] The specific solution process is as follows: The computing device formalizes the problem into a constrained optimization: maximization ,satisfy To address the non-monotonicity and multiple local optima of the function, the computing device employs a sequential quadratic programming algorithm and iteratively solves the problem from multiple initial points. For example, from... From convergence to (Non-performing loan rate approximately 1.50%, loan amount 427,500 yuan), from From convergence to (Fault rate approximately 1.98%, credit limit 576,000 yuan). By comparing the objective function values, the globally optimal adjustment coefficient is determined for the equipment. .

[0149] It should be understood that this implementation transforms the solution for the adjustment coefficient into a constrained mathematical optimization problem. A Lagrangian function is constructed with the defect rate not exceeding a threshold as a safety constraint and maximizing resource allocation as the business objective. Numerical optimization algorithms are then used to solve for the optimal adjustment coefficient. This allows the adjustment coefficient to be chosen in a way that satisfies both safety control requirements and maximizes business growth through resource allocation.

[0150] In some embodiments, because the mapping relationship between group type and adjustment coefficient is generated based on historical data, its effectiveness may drift with changes in external factors such as market environment and user behavior patterns. Therefore, the computing device can periodically or triggerically update and iterate this mapping relationship based on newly generated resource allocation data to ensure that the quota decision-making strategy always remains adapted to the current security features. This process is as follows: Figure 5 As shown, the specific steps include the following: S501-S504: S501. For any third group among different groups, obtain resource allocation data for resource allocation based on the mapping relationship between the third group application group type and the adjustment coefficient.

[0151] Resource allocation data refers to the process and result records generated by the actual resource processing of resource requesters belonging to the third group after applying the currently established mapping relationship between group types and adjustment coefficients.

[0152] Specifically, resource allocation data includes at least the following: resource requester identifier, resource allocation timestamp, application target adjustment coefficient, target resource allocation amount calculated and issued based on the coefficient, and resource return status record corresponding to the allocation. The resource return status record is used to subsequently determine whether the resource request ultimately expires.

[0153] Specifically, in order to assess and maintain the timeliness and effectiveness of the mapping relationship between group types and adjustment coefficients, computing devices need to acquire data samples reflecting the latest business conditions generated after the application of this mapping relationship. Therefore, by collecting resource allocation data for the third group in real time or periodically, computing devices can accumulate the necessary empirical data foundation for subsequent analysis of the group's actual performance under the current strategy (such as calculating the actual defect rate) and business effectiveness.

[0154] One possible implementation involves the computing device querying resource allocation data from business logs or other logs. Specifically, the computing device sets a starting point from the last mapping update or evaluation, and continuously captures or queries all resource allocation records marked as belonging to the third group from the business database, extracting relevant data fields until a preset data volume or time period is reached.

[0155] For example, if the mapping relationship takes effect on January 1, 2023, and the computing device initiates the evaluation of the third group on March 1, 2023, then it will obtain all loan disbursement records and subsequent repayment status records of the third group that were processed based on the mapping relationship from January 1, 2023 to February 28, 2023, which constitute the resource allocation dataset to be analyzed.

[0156] S502. Based on resource allocation data, determine the actual resource request failure rate of the third group.

[0157] The actual resource request failure rate refers to the proportion of resource request failures that have actually occurred, calculated based on the resource allocation data newly generated after the application of the current mapping relationship in the third group.

[0158] One possible implementation involves the computing device executing a structured query statement. The computing device writes a query for the database table storing resource allocation data to count the total number of records belonging to the third group within a specified time range, and simultaneously counts the number of records whose resource return overdue status field meets the "overdue" condition. Finally, the latter is divided by the former and multiplied by 100% to obtain the actual resource request failure rate as a percentage.

[0159] S503. Based on the quantitative relationship between the adjustment coefficient of the third group and the defect rate, and the adjustment coefficient of the third group, determine the expected defect rate of the third group.

[0160] The expected defect rate refers to the theoretical defect rate calculated by substituting the adjustment coefficient used in the current mapping relationship for the third group into the existing adjustment coefficient and defect rate quantification relationship model for that group. This value characterizes the theoretical defect rate level that should be generated by applying the current adjustment coefficient from the perspective of the model fitted to historical data, providing a benchmark for evaluating the accuracy of model predictions.

[0161] Specifically, to evaluate the predictive ability of the existing quantitative relationship model for the current business situation, the computing device needs to use this model to calculate the theoretical defect rate output under the current strategy (i.e., the current adjustment coefficient). Therefore, the computing device uses the adjustment coefficient currently being used by the third group. As input, its corresponding quantization relation function is called and executed. The expected request failure rate for this group can be calculated directly. This allows us to obtain a theoretical prediction that can be compared with the actual observed value.

[0162] S504. If the deviation rate between the actual resource request failure rate and the expected request failure rate of the third group is greater than or equal to the preset deviation rate threshold, the adjustment coefficient of the third group shall be re-determined based on the resource allocation data.

[0163] The deviation rate refers to the degree of deviation between the actual resource request failure rate and the expected request failure rate.

[0164] Specifically, the computing device calculates the deviation rate between the actual resource request failure rate and the expected request failure rate to quantitatively assess the difference between the actual performance resulting from the adjustment coefficient of the current application and the performance control target that the coefficient is expected to achieve. Therefore, when the computing device determines that the deviation rate exceeds a preset threshold, it means that the adjustment coefficient can no longer stably control the performance level within the expected target range under the current business environment. Based on the latest resource allocation data, it then initiates a process to redetermine the adjustment coefficient, generating a new coefficient that matches the current performance characteristics, ensuring that the control effect of the resource allocation strategy meets expectations.

[0165] For example, the preset deviation rate threshold can be set to 20% or 0.15. The deviation rate can be calculated as: Actual Defect Rate Deviation = |Actual Defect Rate - Expected Defect Rate| / Expected Defect Rate. This application embodiment does not limit the value of the preset deviation rate threshold.

[0166] In some embodiments, the computing device may, based on resource allocation data, redetermine the adjustment coefficients of the third group in steps S201-S204.

[0167] In some embodiments, after the computing device redetermines the adjustment coefficients for the third group based on resource allocation data, it may also perform weighted smoothing based on the old and new adjustment coefficients. The computing device can then apply the newly determined adjustment coefficients... The original adjustment coefficient in the mapping relationship Perform a weighted average and obtain the weighted value. As the adjustment coefficient ultimately updated in the mapping relationship, where θ These are preset weight values. This method can prevent coefficients from changing drastically due to short-term data fluctuations, thus improving the stability of the strategy.

[0168] For example, θ The value range can be set to [0.6, 0.8] or [0.7, 0.9], etc. This application embodiment specifies... θ There is no restriction on the range of values ​​that can be obtained.

[0169] It should be understood that this implementation method determines the effectiveness of existing adjustment coefficients by comparing the deviation between the actual and expected defect rates. When the deviation exceeds a threshold, the adjustment coefficients are refitted and determined based on the latest resource allocation data. This ensures that the mapping relationship between group types and adjustment coefficients can adapt to dynamic changes in data distribution, effectively improving the timeliness and adaptability of the adjustment strategy. This ensures that a balance between resource allocation security and business growth can be maintained even when market conditions or user characteristics change.

[0170] In some embodiments, the computing device may further dynamically adjust the preset failure rate threshold based on trend warning after determining the actual resource request failure rate of the third group.

[0171] One possible implementation, after S502, further includes: when the ratio of the actual resource request failure rate of the third group to the preset failure rate threshold corresponding to the third group is greater than or equal to the preset warning threshold, updating the adjustment coefficient of the third group in the mapping relationship between group type and adjustment coefficient based on the preset scaling factor.

[0172] The warning threshold is greater than 1. The preset scaling factor is less than 1.

[0173] For example, the warning threshold can be set to 1.4, 1.5, or 1.6, etc. The preset scaling factor can be set to 0.15, 0.2, or 0.25, etc. The specific values ​​of the warning threshold and the preset scaling factor are not limited in this embodiment.

[0174] One possible implementation involves equipping the computing device with a timed detection module. After each S502 calculation of the actual defect rate, this module calculates its ratio to a preset defect rate threshold and compares it with a warning threshold. If a trigger condition is met, a coefficient update interface is invoked. This interface reads the current adjustment coefficient for that group from the mapping table, multiplies it by a preset scaling factor, obtains a new adjustment coefficient, and writes it directly back to the mapping table, thus completing rapid intervention. The entire process does not rely on complex model refitting, achieving millisecond-level response.

[0175] For example, the preset default rate threshold is 2%, the warning threshold is 1.5, and the preset scaling factor is 0.8. If the calculated actual default rate for a certain group is 3.2%, the ratio is 1.6 (>1.5), triggering a warning. The calculation device multiplies the group's current adjustment coefficient of 1.25 by 0.8 to obtain a new coefficient of 1.0, and immediately updates the mapping relationship. Thereafter, new requests from this group will be processed using coefficient 1.0, and the credit limit will be rapidly reduced.

[0176] It should be understood that this implementation establishes a safety early warning mechanism by setting the ratio of the actual defect rate to a preset threshold as the early warning trigger condition. This mechanism can achieve rapid and conservative adjustments to the resource allocation strategy for this group without relying on complex model reconstruction, so as to suppress the further development of poor performance and thus maintain the safety and controllability of resource allocation.

[0177] like Figure 6 This is a schematic diagram of the structure of a resource processing device provided in an embodiment of this application. Figure 6 As shown, the resource processing device includes: an acquisition module 601 and a processing module 602.

[0178] The acquisition module 601 is used to acquire the multi-dimensional characteristic attributes of the resource requester and the basic resource allocation quota; the multi-dimensional characteristic attributes include: the security verification level of the resource requester, the type of the requested resource, the health of the domain in which the requested resource is applied, and the behavioral characteristics of the resource request.

[0179] The processing module 602 is used to construct a target attribute value combination based on the values ​​of each feature attribute in the multi-dimensional feature attributes, and to determine the group type of the target group to which the resource requester belongs from a preset attribute value combination and group type mapping table; wherein, the attribute value combination and the group type correspond one-to-one; based on the group type of the target group and the mapping relationship between the group type and the adjustment coefficient, the target adjustment coefficient corresponding to the resource requester is determined; the group type of the target group and the target adjustment coefficient correspond one-to-one; the adjustment coefficient is obtained by constraint analysis of the fitting relationship between the adjustment coefficient and the resource request failure rate in the group; the target adjustment coefficient is used to increase the basic resource allocation quota value under the premise of meeting the preset group security conditions; the resource request failure rate is obtained by the number of resource requesters in the corresponding group whose resource return overdue status is overdue; the basic resource allocation quota is adjusted based on the target adjustment coefficient to obtain the target resource allocation quota of the resource requester, and the resource is processed with the target resource allocation quota.

[0180] In other embodiments, the processing module 602 is specifically used to acquire historical resource allocation data for different groups. This historical resource allocation data includes resource allocation records for different historical resource requesters, and each record includes historical resource allocation amounts and overdue resource return status. Feature fitting processing is performed on the historical resource allocation data for each group to obtain a quantitative relationship between the adjustment coefficient and the failure rate for each group. Based on this quantitative relationship, and with the expectation that the expected resource request failure rate is less than or equal to a preset failure rate threshold for the corresponding group as a constraint, the adjustment coefficient for each group is determined. The expected resource request failure rate is obtained by mapping the adjustment coefficient to be determined through the quantitative relationship. A mapping relationship between group type and adjustment coefficient is constructed based on the group type and the corresponding adjustment coefficient for each group.

[0181] In other embodiments, processing module 602, for any first group among different groups, performs logistic regression fitting based on the historical resource allocation data corresponding to the first group, using historical resource allocation amount as a feature and overdue resource return status as a label, to obtain a logistic regression function. The logistic regression function represents the quantitative relationship between historical resource allocation amount and resource request failure rate. The logistic regression function includes: feature weight coefficients and an intercept term. The feature weight coefficients represent the degree of influence of resource allocation amount on resource request failure rate within the first group. The intercept term represents the degree of influence of non-resource allocation amount factors on resource request failure rate within the first group. The average resource allocation amount of the first group is determined. Using the amount adjusted based on the adjustment coefficient to be determined as the independent variable of the logistic regression function, the quantitative relationship between the adjustment coefficient and failure rate corresponding to the first group is obtained.

[0182] In other embodiments, the preset group security condition includes: the expected resource request failure rate corresponding to the target group is less than or equal to the preset failure rate threshold of the target group.

[0183] In other embodiments, the processing module 602 is specifically used to construct a Lagrangian function corresponding to any second group among different groups, with the constraint that the expected resource request failure rate is less than or equal to the preset failure rate threshold of the corresponding group, and with the objective of maximizing the resource allocation amount. Based on numerical optimization algorithms and the Lagrangian function, the adjustment coefficient of the second group is determined.

[0184] In other embodiments, the processing module 602 is further configured to acquire resource allocation data for any third group among different groups, based on the mapping relationship between the application group type and the adjustment coefficient of the third group, and to allocate resources accordingly. Based on the resource allocation data, the actual resource request failure rate of the third group is determined. Based on the quantitative relationship between the adjustment coefficient and the failure rate of the third group, and the adjustment coefficient of the third group, the expected request failure rate of the third group is determined. If the deviation rate between the actual resource request failure rate and the expected request failure rate of the third group is greater than or equal to a preset deviation rate threshold, the adjustment coefficient of the third group is re-determined based on the resource allocation data.

[0185] In other embodiments, the processing module 602 is further configured to, when the ratio of the actual resource request failure rate of the third group to the preset failure rate threshold corresponding to the third group is greater than or equal to a preset warning threshold, update the adjustment coefficient of the third group in the mapping relationship between group type and adjustment coefficient based on a preset scaling factor. The warning threshold is greater than 1. The preset scaling factor is less than 1.

[0186] The resource processing apparatus 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.

[0187] Figure 7 This is a schematic diagram of the structure of a resource processing device provided in an embodiment of this application. Figure 7 As shown, the resource processing device includes: a memory 701, a transceiver 702, and at least one processor 703.

[0188] The transceiver 702 is used to interact with other devices to send and receive data.

[0189] For example, in this embodiment of the application, transceiver 702 can be used to obtain the multi-dimensional characteristic attributes of the resource requester and the basic resource allocation quota.

[0190] The memory 701 is used to store computer program code, which includes computer instructions. These computer instructions run in the resource processing device described above to implement the method shown in the above 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.

[0191] Processor 703 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 703 can also be other general-purpose processors. The general-purpose processor can be a microprocessor or any conventional processor.

[0192] The memory 701, transceiver 702, and processor 703 are communicatively connected. For example, the memory 701 and transceiver 702 can be connected to the processor 703 via a system bus to complete communication between them. 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.

[0193] Optionally, the memory 701 can be either standalone or integrated with the processor 703. When the memory 701 is set up independently, it is connected to the processor 703 via a system bus.

[0194] This application also provides a chip for executing instructions, which is used to execute the resource processing method described in the above embodiments.

[0195] 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 resource processing method described in the above embodiments. Specifically, when the computer instructions are executed by a processor, the resource processing device can execute the technical solution of the resource processing method described in the above embodiments.

[0196] 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, and when the at least one processor executes the computer program, it can implement the technical solution of the resource processing method in the above embodiments.

[0197] 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 medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0198] 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 application-specific integrated circuits (ASICs). 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.

[0199] 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.

[0200] 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.

[0201] 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.

[0202] 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.

[0203] 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.

[0204] 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.

[0205] 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 resource processing method, characterized in that, include: Obtain the multidimensional characteristic attributes of the resource requester and the basic resource allocation quota; The multidimensional feature attributes include: the security verification level of the resource requester, the type of the requested resource, the health of the domain in which the requested resource is applied, and the behavioral characteristics of the resource request. Based on the values ​​of each of the multidimensional feature attributes, a target attribute value combination is constructed. From a preset attribute value combination and group type mapping table, the group type of the target group to which the resource requester belongs is determined; wherein, the attribute value combination and the group type correspond one-to-one. Based on the group type of the target group and the mapping relationship between the group type and the adjustment coefficient, the target adjustment coefficient corresponding to the resource requester is determined; the group type of the target group corresponds one-to-one with the target adjustment coefficient; the adjustment coefficient is obtained by constraint analysis of the fitting relationship between the adjustment coefficient and the resource request failure rate in the group; the target adjustment coefficient is used to increase the value of the basic resource allocation quota under the premise of meeting the preset group security conditions; the resource request failure rate is obtained by the number of resource requesters in the corresponding group whose resource return overdue status is overdue; The basic resource allocation quota is adjusted based on the target adjustment coefficient to obtain the target resource allocation quota for the resource requester, and resource processing is performed using the target resource allocation quota.

2. The method according to claim 1, characterized in that, The mapping relationship between the group type and the adjustment coefficient is obtained in the following way: Obtain historical resource allocation data for each of the different groups; the historical resource allocation data includes resource allocation records for different historical resource requesters; The resource allocation record includes: historical resource allocation amount and overdue resource return status; Feature fitting processing is performed on the historical resource allocation data corresponding to each group to obtain the quantitative relationship between the adjustment coefficient and the defect rate of each group; Based on the quantitative relationship between the adjustment coefficient and the defect rate of each group, and with the constraint that the expected resource request defect rate is less than or equal to the preset defect rate threshold of the corresponding group, the adjustment coefficient corresponding to each group is determined; the expected resource request defect rate refers to the rate obtained by mapping the adjustment coefficient to be determined through the quantitative relationship. Based on the group type of each group and the adjustment coefficient corresponding to each group, a mapping relationship between the group type and the adjustment coefficient is constructed.

3. The method according to claim 2, characterized in that, The feature fitting includes logistic regression fitting; The step of performing feature fitting processing on the historical resource allocation data corresponding to each group to obtain the quantitative relationship between the adjustment coefficient and the defect rate of each group includes: For any one of the different groups, based on the historical resource allocation data corresponding to the first group, using the historical resource allocation amount as a feature and the overdue resource return status as a label, a logistic regression fit is performed to obtain a logistic regression function; wherein, the logistic regression function represents the quantitative relationship between the historical resource allocation amount and the resource request failure rate; the logistic regression function includes: feature weight coefficients and an intercept term; the feature weight coefficients represent the degree of influence of the resource allocation amount on the resource request failure rate within the first group; the intercept term represents the degree of influence of non-resource allocation amount factors on the resource request failure rate within the first group; Determine the average resource allocation amount for the first group; Using the adjusted amount of the average resource allocation based on the adjustment coefficient to be determined as the independent variable of the logistic regression function, the quantitative relationship between the adjustment coefficient and the non-performing rate corresponding to the first group is obtained.

4. The method according to claim 2, characterized in that, The preset group security conditions include: the expected resource request failure rate corresponding to the target group is less than or equal to the preset failure rate threshold of the target group.

5. The method according to claim 2, characterized in that, The quantitative relationship between the adjustment coefficient and the defect rate for each group, with the expected resource request defect rate being less than or equal to a preset defect rate threshold for the corresponding group as a constraint, determines the adjustment coefficient corresponding to each group, including: For any second group among the different groups, with the expected resource request failure rate being less than or equal to the preset failure rate threshold of the corresponding group as a constraint, and with the goal of maximizing the resource allocation amount, a Lagrange function corresponding to the second group is constructed. The adjustment coefficients for the second group are determined based on the numerical optimization algorithm and the Lagrange function.

6. The method according to claim 2, characterized in that, The method further includes: For any third group among the different groups, obtain the resource allocation data of the third group that uses the mapping relationship between the group type and the adjustment coefficient to allocate resources; Based on the resource allocation data, the actual resource request failure rate of the third group is determined; Based on the quantitative relationship between the adjustment coefficient and the defect rate of the third group, and the adjustment coefficient of the third group, the expected requested defect rate of the third group is determined. If the deviation rate between the actual resource request failure rate and the expected resource request failure rate of the third group is greater than or equal to a preset deviation rate threshold, the adjustment coefficient of the third group is re-determined based on the resource allocation data.

7. The method according to claim 6, characterized in that, After determining the actual resource request failure rate of the third group based on the resource allocation data, the method further includes: If the ratio of the actual resource request failure rate of the third group to the preset failure rate threshold corresponding to the third group is greater than or equal to the preset warning threshold, the adjustment coefficient of the third group in the mapping relationship between the group type and the adjustment coefficient is updated based on the preset scaling factor; the warning threshold is greater than 1; and the preset scaling factor is less than 1.

8. A resource processing device, characterized in that, include: The acquisition module is used to acquire the multi-dimensional characteristic attributes of the resource requester and the basic resource allocation quota; The multidimensional feature attributes include: the security verification level of the resource requester, the type of the requested resource, the health of the domain in which the requested resource is applied, and the behavioral characteristics of the resource request. The processing module is used to construct target attribute value combinations based on the values ​​of each of the multidimensional feature attributes; determine the group type of the target group to which the resource requester belongs from a preset attribute value combination and group type mapping table; wherein, the attribute value combinations and group types correspond one-to-one; determine the target adjustment coefficient corresponding to the resource requester based on the group type of the target group and the mapping relationship between the group type and the adjustment coefficient; the group type of the target group and the target adjustment coefficient correspond one-to-one; the adjustment coefficient is obtained by constraint analysis of the fitting relationship between the adjustment coefficient and the resource request failure rate in the group; the target adjustment coefficient is used to increase the value of the basic resource allocation quota under the premise of meeting the preset group security conditions; the resource request failure rate is obtained by the number of resource requesters in the corresponding group whose resource return overdue status is overdue; adjust the basic resource allocation quota based on the target adjustment coefficient to obtain the target resource allocation quota of the resource requester, and process resources with the target resource allocation quota.

9. A resource processing device, characterized in that, include: 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 resource processing device performs the method as described in any one of claims 1-7.

10. 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 method as described in any one of claims 1-7.

11. A computer program product, characterized in that, When the computer program product is run on a computer / executed by the computer's processor, it implements the method as described in any one of claims 1-7.

Images