Assembly line design method based on small enterprise credit factory mode

Abstract

The invention relates to the technical field of data processing, and discloses an assembly line design method based on a small enterprise credit factory mode, which comprises the following steps of: initializing task and resource parameters, collecting a credit approval task set and an approval role set, defining processing time, priority weight and latest completion time for each task, and completing the task according to the processing time, the priority weight and the latest completion time. Defining a resource capability and an initial load for each approval role; and constructing a multi-objective optimization model, comprehensively considering task completion time, a resource utilization rate and task priority completion quality, and converting an optimization objective combination weight into a comprehensive objective function. According to the invention, a technical scheme based on customer admission management and detailed approval paths is adopted, and precision of risk assessment and hierarchical management of loan customers is realized. The approval nodes can be dynamically matched according to the risk levels of the customers, and the problem of resource waste caused by homogenization processing of high-risk customers and low-risk customers in an existing process is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and specifically to a pipeline design method based on the credit factory model for small enterprise credit. Background Art

[0002] With the rapid development of the credit business for small and micro enterprises, the efficiency of the loan approval process and the risk control ability have become the key issues restricting the improvement of the service capabilities of financial institutions. The traditional credit approval process usually adopts a static process design, where each link is executed in a fixed order, and it is difficult to make dynamic adjustments according to the customer risk level and real-time business requirements. This method not only prolongs the approval time but also restricts the effective utilization of resources; The prior art has proposed some methods for fixed approval processes, which process credit tasks through preset approval nodes and fixed resource allocation mechanisms. These technologies usually rely on static allocation models, allocate all tasks to approval roles according to preset rules, and complete the work through established approval chains. After the approval roles gradually process the tasks, the results are passed to the next node until the final approval is completed. Although such technologies can meet the basic approval requirements, their flexibility is insufficient in scenarios with increased task complexity and diversification; However, the existing approval processes have significant deficiencies in dealing with the dynamic changes of the credit business for small and micro enterprises. The approval process is rigid and cannot dynamically adjust the approval path according to the customer risk level. There is no difference in the processing methods for high-risk and low-risk customers, resulting in unreasonable resource allocation. Secondly, there is a lack of a dynamic optimization mechanism for task allocation, and the load of approval roles is severely uneven. Some roles are overloaded, while other roles are in a low-load state. In addition, the information transmission and collaboration mechanisms are not perfect, resulting in frequent information delays and duplicate task processing in the approval chain, and the approval efficiency is low. Finally, the existing technologies cannot monitor the key indicators in the approval process in real time, lack dynamic feedback and optimization means for parameters such as queue length and resource utilization rate, and it is difficult to adapt to complex and changeable approval scenarios. These deficiencies significantly reduce the efficiency and quality of credit approval, and an innovative method is needed to improve and optimize the existing process. Summary of the Invention

[0003] Aiming at the deficiencies of the prior art, the present invention provides a pipeline design method based on the credit factory model for small enterprise credit, which solves the problems of the existing approval process being rigid, unable to dynamically adjust the path, uneven task allocation, lagging information transmission, lack of real-time monitoring and optimization means, difficult to adapt to complex scenarios, and limited efficiency and quality.

[0004] To achieve the above objectives, the present invention is realized through the following technical solutions: A pipeline design method based on the credit factory model for small enterprise credit, including the following steps: Initialize the task and resource parameters, collect the credit approval task set and the approval role set, define the processing time, priority weight, and latest completion time for each task, and define the resource capacity and initial load for each approval role; Construct a multi-objective optimization model, comprehensively consider the task completion time, resource utilization rate, and task priority completion quality, and transform the optimization objectives combined with weights into a comprehensive objective function; Use an optimization algorithm to solve the comprehensive objective function, generate a task allocation plan, and complete the initial task allocation based on the constraint conditions; Dynamically adjust the task allocation plan. When a new task arrives, calculate the revenue function of each approval role by constructing a dynamic game model, and optimize and solve its task allocation strategy; Real-time monitor the approval queue length, resource utilization rate, and task completion time, dynamically adjust the optimization weights according to the monitoring results, and further optimize the task allocation plan; Output the final task allocation results, including approval efficiency, resource utilization rate, and priority completion status.

[0005] Preferably, the optimization objectives include: Minimize the total completion time of the credit approval pipeline by calculating the allocation time of all tasks among different approval roles; Maximize the resource utilization rate of the approval role by increasing the proportion of the task load assigned to the approval role to its resource capacity; Maximize the task priority completion quality by preferentially processing tasks with higher priorities and increasing the completion ratio of high-priority tasks.

[0006] Preferably, the multi-objective optimization model includes the following constraint conditions: Each task must be assigned to one and only one approval role; The total load of the approval role shall not exceed its resource capacity; The completion time of each task shall not exceed its latest completion time.

[0007] Preferably, the non-dominated sorting genetic algorithm is used to solve the comprehensive objective function, including the following steps: Initialize the population, generate multiple individuals according to the random task allocation plan, and each individual is a task allocation matrix X = [x i,j , where x i,j ∈ {0, 1} indicates whether task t i is assigned to approval role r j ; Based on the comprehensive objective function F, perform non-dominated sorting on each individual in the population. The comprehensive objective function is defined as: Among them, C i,j is the task t i in the approval role rj Processing time; Capacity(r j ): Approval role r j Resource capacity; P i : Task t i Priority weight; w 1 , w 2 , w 3 is the weight coefficient; C i,j is the expected processing time of task t i on approval role r j ; x i,j indicates whether task t i is assigned to approval role r j ; n is the total number of tasks; m is the total number of approval roles; Capacity(r j ) is the maximum resource capacity of approval role r j ; P i Task t i Priority weight; Perform genetic operations, including selection, crossover, and mutation, to generate a new generation of population: Selection operation: Select individuals according to the fitness value with probability; Crossover operation: Randomly exchange the task assignment matrices of the selected individuals to generate a new task assignment scheme; Mutation operation: Randomly adjust the value of x i,j in the task assignment matrix to explore a new solution space; Iterate cyclically until the population converges to obtain the Pareto optimal solution set; Select a solution that suits the current constraint conditions from the Pareto optimal solution set as the initial task assignment scheme.

[0008] Preferably, the task assignment scheme includes: Based on the dynamic game model, regard the approval role as a game participant, and construct a revenue function through the quality of task priority completion and resource utilization rate; Determine the optimal task assignment strategy for each approval role by optimizing and solving the Nash equilibrium of the game model.

[0009] Preferably, the function of the dynamic game model is expressed by the following formula: The revenue function U j of approval role r j consists of two parts: the quality of task priority completion and resource utilization rate, and is defined as: Among them, T j is the set of tasks assigned to role r j ; P iFor task t i is the priority weight; Capacity(r j ) is the resource capacity of approval role r j ; L j is the current load of approval role r j ; α, β are weight parameters used to balance the impact of task priority completion quality and resource utilization; The goal of the game model is to make the benefits of all approval roles reach Nash equilibrium, that is, the benefit of each role reaches the maximum when the strategies of other roles are fixed, and the following conditions are satisfied: Among them, represents the partial derivative of the benefit function U j with respect to the strategy of role r j ; The Nash equilibrium solution is achieved through the following steps: Initialize the task assignment scheme T j ; For each role r j , optimize the benefit function U j while fixing the strategies of other roles; Iteratively update the assignment scheme until the benefit functions of all roles converge to the equilibrium state.

[0010] Preferably, the real-time monitoring includes: Monitoring the length of the approval queue to judge the task backlog situation; Monitoring the resource utilization rate of the approval role to optimize task assignment; Adjusting and optimizing the weights according to the monitoring results to cope with the approval peak.

[0011] Preferably, the adjustment of the optimized weights satisfies the following conditions: When the length of the approval queue exceeds the preset threshold, increase the weight of approval efficiency; When the resource utilization rate is unbalanced, increase the weight of resource utilization rate; When the completion rate of priority tasks is low, increase the weight of priority task completion quality.

[0012] Preferably, the assignment result includes: The final task assignment scheme, including the approval role assignment information of each task; The key performance indicators of the credit approval pipeline, including the total approval time, resource utilization rate, and the completion of priority tasks.

[0013] The present invention also provides a pipeline design system based on the credit factory model for small enterprises, including: A data collection module for collecting relevant data on the credit approval task set and the approval role set; An optimization calculation module for constructing an optimization model and solving the comprehensive objective function; A dynamic scheduling module for adjusting the task allocation scheme based on the dynamic game model; A monitoring and feedback module for real-time monitoring of the approval queue, resource utilization rate, and task completion status, and dynamically adjusting the optimization weights; A data output module for outputting the final task allocation result and key performance indicators.

[0014] The present invention provides a pipeline design method based on the credit factory model for small enterprise credit. It has the following beneficial effects: 1. The present invention adopts a technical solution based on customer access management and refined approval paths, achieving precision in the risk assessment and hierarchical management of loan customers. Compared with the technical solution in the prior art where a single approval process lacks the ability of dynamic adjustment, the present invention can dynamically match approval nodes according to the risk level of customers, solving the deficiency of resource waste caused by the homogeneous treatment of high-risk and low-risk customers in the existing process.

[0015] 2. The present invention introduces a multi-stage division and conditional diversion mechanism for the approval process, enabling the clear definition of responsibilities and risk control points in each stage. Compared with the technical solution in the prior art where the approval chain is long and the responsibilities are unclear, the present invention effectively reduces process duplication and information transmission errors, solving the problems of long approval time and insufficient control of key nodes, and greatly improving the approval efficiency.

[0016] 3. The present invention, through the design scheme of dynamic task allocation and approval role matching, enables the balanced allocation of loads at different approval nodes. Compared with the technical solution in the prior art where the utilization rate of approval resources is unbalanced and some roles are overloaded, the present invention significantly improves the resource utilization efficiency, solving the deficiency of approval bottlenecks caused by unreasonable task allocation in the existing process.

[0017] 4. The present invention adopts a real-time monitoring and feedback adjustment mechanism, dynamically optimizing the task allocation path in combination with task status data, achieving the technical effect of continuously optimizing the approval process. Compared with the technical solution in the prior art where the process is fixed and lacks a flexible adjustment mechanism, the present invention is more adaptable in dealing with changes in customer needs, solving the deficiency of the existing system's difficulty in responding to the dynamic approval environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is the flowchart of the method steps of the present invention; Figure 2 is the system architecture diagram of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] Please refer to the attached Figure 1 , the embodiment of the present invention provides a pipeline design method based on the credit factory model for small enterprise credit. Through the construction of a multi-objective optimization model and the implementation of a dynamic task allocation strategy, including optimizing approval efficiency, balancing resource utilization, and ensuring the completion rate of high-priority tasks, a comprehensive optimization of the pipeline design based on the credit factory model for small enterprise credit is achieved. Furthermore, the overall approval efficiency, resource usage rationality, and response ability to high-priority tasks are improved, including the following steps: S1. Initialize task and resource parameters; S2. Construct a multi-objective optimization model; S3. Solve the multi-objective optimization and perform initial task allocation; S4. Dynamically adjust the task allocation plan; S5. Monitor in real time and make feedback adjustments; S6. Output results and evaluate the system.

[0021] For step S1, in this embodiment, the task set is denoted as T = {t 1 , t 2 , t n}. Wherein: t i represents a specific approval task, i = 1, 2, n, and n is the number of tasks in the current system.

[0022] Each task t i has the following attributes: Processing time C i,j : The estimated processing time of task t i on the approval role r j , in minutes, and the specific value is determined according to historical approval data or business rules; Priority weight P i : The importance of task t i , with a value range of [0, 1]. The larger the priority weight, the more important the task; Latest completion time T deadline : The time limit by which task t i must be completed, usually recorded in absolute time format (such as YYYY-MM-DD HH:MM:SS).

[0023] As an option, a task set T can be generated in real time based on an enterprise management system. In this implementation, the task attributes C i,j , P i and T deadline can be imported from an external database.

[0024] Specifically, the calculation of the processing time C i,j can be combined with the following formula: Where: W i represents the workload of task t i , in standard task units; E j represents the unit processing capacity of the approval role r j , in standard task units per minute.

[0025] The setting of the priority weight P i is predefined according to business rules. For example, the priority weight of a high-risk loan task is usually greater than that of an ordinary loan task.

[0026] In some embodiments, to enhance flexibility, task label information can be added during task initialization for classified management of tasks. For example, "credit loan" tasks and "guaranteed loan" tasks can be labeled with different tags for subsequent optimized model differentiation processing.

[0027] The set of approval roles is denoted as R = {r 1 , r 2 , r m}. Where: r j represents a specific approval role, j = 1, 2, m, where m is the number of approval roles in the current system; Each approval role r j has the following attributes: Resource capacity Capacity(r j ): The maximum task processing capacity of the approval role r j per unit time, in standard task units per minute; Current load L j (0): The current task load of the approval role r j , with an initial value of 0, and will be dynamically updated according to the task allocation scheme later.

[0028] Generally, the setting of the resource capacity Capacity(r j ) is related to factors such as the position level and work experience of the approval role. For example, the task processing capacity of a branch manager is usually higher than that of an ordinary customer manager.

[0029] In a possible implementation, the resource capacity of the approval role can be estimated by the following formula: Capacity(r j )=B j ·H j Where B j represents the basic processing capacity of the approval role r j , in units of standard task units per hour; H j represents the daily working hours of the approval role, in hours.

[0030] Specifically, the resource capacity can be obtained through historical data statistics. For example, based on the approval data of the past month, calculate the average processing capacity of each role.

[0031] After initialization is completed, the attribute data of the task set T and the approval role set R need to be stored in a unified data structure. As an implementation, a multi-dimensional array or an associative mapping table can be used for storage. For example: Storage structure of the task set T: Primary key: task identifier t i ; Attribute fields: processing time C i,j , priority weight P i , latest completion time T deadline .

[0032] Storage structure of the approval role set R: Primary key: role identifier r j ; Attribute fields: resource capacity Capacity(r j ), current load L j (0).

[0033] In some embodiments, the system can also dynamically adjust the attributes of the task set and the approval role set. For example, when a new approval role joins, the data structure of the set R can be automatically updated.

[0034] To ensure the accuracy and integrity of the initialization data, the present invention designs a data consistency verification mechanism, which specifically includes: Verify the integrity of the task set T to ensure that the processing time, priority weight, and completion deadline of each task are defined; Verify the rationality of the approval role set R to ensure that the resource capacity and initial load of each role are in line with the actual situation; Verify the matching of the task set and the role set to ensure that the processing time of all tasks is within the defined range of the role resource capacity.

[0035] As an extended function, the present invention supports the dynamic update of the task set. For example, when the system receives a new task, it can automatically call the task initialization module to add the new task t new to the task set T. At the same time, the load of the role set R is recalculated according to the attributes of the new task, and the resource allocation scheme is dynamically adjusted.

[0036] For step S2, by comprehensively considering the task completion time, resource utilization rate, and task priority completion quality, the optimization objectives and constraint conditions are established, providing a theoretical basis for subsequent optimization solving and task allocation. Generally, the objectives of credit approval are diverse, involving multiple aspects such as approval efficiency, resource balance, and priority guarantee. These objectives often conflict with each other. The present invention uniformly describes each objective through mathematical modeling and realizes weight adjustment in the optimization model, making it flexible and adaptable in practical applications.

[0037] In this embodiment, the optimization objectives of the credit approval task are designed into three core directions: minimizing the approval completion time, maximizing the resource utilization rate, and maximizing the completion rate of high-priority tasks. These three objectives are expressed through a comprehensive objective function. Specifically: First, to minimize the approval completion time, the total approval time is defined as: where n represents the total number of tasks; m represents the total number of approval roles; C i,j represents the processing time of task t i on the approval role r j , in minutes; x i,j is a binary decision variable. When task t i is assigned to the approval role r j , x i,j = 1, otherwise 0.

[0038] Secondly, to improve the resource utilization rate of the approval role, the resource utilization rate objective is defined as: where Capacity(r j ) represents the maximum resource capacity of the approval role r j , in standard task units / minute; the numerator part represents the actual assigned task load of the approval role r j .

[0039] To ensure the completion rate of high-priority tasks, the priority objective is defined as: where P i represents task ti The priority weight ranges from [0, 1]; higher-priority tasks correspond to greater weights, and the priority weight is determined by the business rules of credit approval. n is the total number of tasks, and m is the total number of approval roles.

[0040] To comprehensively consider the above three objectives, the present invention constructs a comprehensive objective function through a weighted combination method: F = w 1 T total - w 2 U + w 3 Q Where: w 1 , w 2 , w 3 Are sub-weight parameters, corresponding to the objectives of approval time, resource utilization rate, and priority completion rate respectively; T total Is the total approval completion time, U is the resource utilization rate, Q is the priority task completion rate, and the sum of the weight parameters satisfies w 1 + w 2 + w 3 = 1, and the specific values are determined by actual requirements.

[0041] In the optimization model, to ensure the rationality of the task assignment scheme, the present invention sets the following constraint conditions: Each task must be assigned to one and only one approval role: The total load of the approval role cannot exceed its resource capacity: The completion time of each task cannot exceed its latest completion time: Where, T is the task set, x i,j Indicates whether task t i Is assigned to approval role r j , T i,start And T i,end Respectively represent the start time and completion time of task t i , m is the total number of approval roles, R is the approval role set, C i,j Is the estimated processing time of task t i On approval role r j , Capacity(r j ) The maximum resource capacity of approval role r j , i is the identifier of the task, T deadline Task t i 's latest completion time.

[0042] In some embodiments, the time window of a task can be further refined. For example, stricter time limits can be set for high-priority tasks.

[0043] As an option, the multi-objective optimization model of the present invention supports dynamic adjustment of weight parameters. For example: During the peak approval period, the weight of w 1 can be increased to prioritize the optimization of the approval completion time; In the case of uneven resource distribution, the weight of w 2 can be increased to balance resource utilization; When high-priority tasks pile up, the weight of w 3 can be increased to ensure that important tasks are completed first.

[0044] Specifically, the rule for dynamically adjusting weight parameters can be set in combination with real-time monitoring data. For example, it can be dynamically updated according to the approval queue length Q(t) and resource utilization U j (t).

[0045] For step S3, an initial task assignment plan is generated based on the optimization results. Generally, since multi-objective optimization involves multiple conflicting objectives, such as minimizing approval time, maximizing resource utilization, and maximizing the completion rate of priority tasks, it is difficult for traditional optimization methods to quickly find the global optimal solution in the solution space. The present invention solves the comprehensive objective function by introducing the non-dominated sorting genetic algorithm (NSGA-II), generates a set of Pareto optimal solutions, and selects a suitable assignment plan from them as the initial task assignment plan.

[0046] In this embodiment, the non-dominated sorting genetic algorithm is designed as the core algorithm for solving multi-objective optimization problems, specifically including population initialization, non-dominated sorting, genetic operations, and solution convergence. The specific process is as follows: Population initialization is the first step. Generate an initial population P 0 , and each individual X k in the population represents a task assignment plan. Specifically, the size N of the population is usually determined according to computing resources, and the encoding form of the individual X k is a task assignment matrix: where x i,j represents whether task t i is assigned to the approval role r j . When x i,j = 1, task t i is assigned to role r j ; n is the total number of tasks; m is the total number of approval roles.

[0047] Generally, the generation of the initial population adopts a randomization strategy to ensure population diversity and cover a large solution space.

[0048] Non-dominated sorting is the second step. For each individual in the population, it is sorted according to the value of the comprehensive objective function F. Specifically, calculate the fitness of each individual on the multi-objective function, including the approval time T total , resource utilization rate U, and priority completion rate Q. The fitness calculation formula is: F(X k ) = w 1 T total - w 2 U + w 3 Q where, T total represents the total approval completion time; U represents the resource utilization rate; Q represents the priority completion rate, and w 1 , w 2 , w 3 are sub-weight parameters.

[0049] Non-dominated sorting will generate different sorting levels according to the objective function values of individuals, and non-dominated solutions are preferentially retained, that is, those individuals that are not inferior to other solutions in all objectives.

[0050] Genetic operations are a crucial part of the optimization process, including selection, crossover, and mutation. The specific process is as follows: In the selection operation, high-quality individuals are selected from the current population through a probability selection method based on fitness to form the parent individuals of the next generation population. As a selection method, the roulette wheel selection method can be used to ensure that individuals with high fitness have a higher probability of being selected.

[0051] In the crossover operation, two individuals are randomly selected from the parent individuals, and part of the task assignment matrix X k is exchanged to generate two new offspring individuals. For example, through the single-point crossover method, the task matrix is cut along a certain column, and the two parts of data are exchanged.

[0052] In the mutation operation, the task assignment matrix X k of the offspring individuals is randomly adjusted with a certain probability. For example, the values of some x i,j in the matrix are changed from 0 to 1, or from 1 to 0. Specifically, the mutation probability is usually set to 1 / n to balance population diversity and convergence speed.

[0053] After the genetic operations are completed, the process of generating the next generation population and non-dominated sorting is entered. Through multiple generations of iteration, the individuals in the population gradually approach the Pareto front. Specifically, when the population converges, the solution set composed of all non-dominated individuals is regarded as the Pareto optimal solution.

[0054] In some embodiments, to ensure the diversity of solutions, a crowding distance calculation method can be introduced. Specifically, calculate the density of each individual in the solution space, and preferentially retain individuals with a larger crowding distance. For step S4, generally, the credit approval task has the characteristics of dynamic changes, such as the real-time arrival of tasks, the load fluctuations of approval roles, etc. Therefore, to cope with these dynamic factors, the present invention introduces a dynamic game model, which uses the game relationship between approval roles to adjust the task allocation scheme in real time to ensure the flexibility and optimality of the allocation scheme.

[0055] The dynamic game model optimizes the quality of task priority completion and resource utilization rate by constructing a revenue function. In practical applications, this model can dynamically adjust the task allocation strategy according to the real-time status of tasks and roles, avoiding performance degradation caused by system fluctuations.

[0056] In this embodiment, the core of the dynamic game model lies in quantifying the revenue of each approval role. The revenue function U j is designed as a weighted combination of two parts, namely the quality of task priority completion and resource utilization rate, and the specific expression is as follows: where U j represents the revenue of approval role r j ; T j is the set of tasks currently assigned to role r j ; P i is the priority weight of task t i , and the value range is [0,1]; Capacity(r j ) is the resource capacity of role r j , with the unit of standard task unit / minute; L j is the current load of role r j , indicating the total processing time of tasks assigned to the role per unit time; α and β are weight parameters used to balance the impacts of the quality of task priority completion and resource utilization rate.

[0057] Generally, the values of the weight parameters α and β can be dynamically adjusted according to system requirements. For example, when the system needs to give priority to completing high-priority tasks, the value of α can be appropriately increased; when the resource utilization rate is unbalanced, the value of β can be appropriately increased.

[0058] In this embodiment, to ensure the rationality and efficiency of task allocation, the game model is optimized with the Nash equilibrium as the goal. Specifically, the Nash equilibrium refers to a state where when the strategies of all approval roles are fixed, any role adjusting its strategy cannot improve its own revenue.

[0059] Specifically, the Nash equilibrium condition of the game model can be expressed as: Among them, represents the revenue function U j with respect to the partial derivative of the strategy of role r j ; R is the set of approval roles.

[0060] In a possible implementation, the solution of the Nash equilibrium is achieved through the following process: Initialize the task assignment scheme T for all roles j ; For each role r j , under the condition of fixing the strategies of other roles, determine the optimal task set T by optimizing the revenue function U j ; Iteratively update the task assignment scheme until the revenue functions of all roles converge to a stable state. j

[0061] In this embodiment, when a new task t new arrives, the system first calculates the attributes of the task, including the processing time C new,j , the priority weight P new and the deadline T deadline,new . Then, according to the revenue function of the current approval role, select the role with the greatest increase in revenue for task assignment.

[0062] As an option, a task reallocation mechanism can be introduced during the dynamic adjustment process, that is, according to the calculation results of the new revenue function, re-adjust the existing task assignment scheme. Specifically, the core of task reallocation lies in comparing the total revenue differences between the existing assignment scheme and the new assignment scheme, and aiming to maximize the total revenue of the system.

[0063] In some embodiments, the dynamic adjustment process can also be combined with real-time monitoring data for weight adjustment. For example, when the system detects that the load of a certain approval role is significantly higher than that of other roles, the value of β can be dynamically increased to reduce the revenue of the high-load role, thereby guiding the system to assign tasks to low-load roles.

[0064] In addition, the dynamic game model of the present invention also supports distributed computing, that is, in a multi-node environment, each node independently calculates its own revenue function and shares the assignment scheme, further improving the task adjustment efficiency.

[0065] For step S5, a real-time monitoring and feedback adjustment mechanism is introduced. Generally, there are various dynamic problems in the credit approval process, such as the length of the approval queue, uneven resource utilization, and accumulation of high-priority tasks. Through the real-time monitoring of key operating parameters, potential bottlenecks can be discovered in a timely manner, and the optimization weights and assignment strategies can be adjusted in combination with the feedback mechanism to maintain the stability and efficiency of the system.

[0066] The real-time monitoring and feedback mechanism of the present invention mainly focuses on three major indicators: the length of the approval queue, resource utilization rate, and task completion status, and realizes closed-loop optimization by dynamically adjusting weights and allocation strategies.

[0067] In this embodiment, the core indicators of real-time monitoring include the length of the approval queue Q(t), the resource utilization rate U j (t) of the approval role, and the completion rate R p (t) of high-priority tasks. The specific calculation methods are as follows: The length of the approval queue Q(t) represents the number of tasks that have not been completed in the current system and is defined as: where n is the total number of tasks; δ i (t) is the status indicator variable of task t i , when t i has not been completed, δ i (t) = 1, otherwise δ i (t) = 0.

[0068] The resource utilization rate U j (t) is used to evaluate the resource load of the approval role r j and is defined as: where L j (t) is the current load of role r j , that is, the total processing time of the tasks assigned to this role; Capacity(r j ) is the maximum resource capacity of role r j .

[0069] The completion rate R p (t) of high-priority tasks reflects the performance of the system in processing high-priority tasks and is defined as: where T p is the set of all high-priority tasks; is the completion status indicator variable of task t i , when t i has been completed, otherwise it is 0; |T p | is the total number of high-priority tasks.

[0070] Generally, the calculation of these indicators can be updated in real time by the system and a complete monitoring report can be generated in a short time.

[0071] In this embodiment, the feedback adjustment mechanism optimizes the weight w by dynamically adjusting1 , w 2 , w 3 to achieve real-time optimization of the allocation strategy. The adjustment rules are as follows: When the length of the approval queue Q(t) exceeds the preset threshold, the system preferentially adjusts the weight w 1 , to increase the priority of the approval efficiency goal and ensure that the task completion time T total is shortened as much as possible.

[0072] When the resource utilization rate U j (t) varies significantly among different roles, for example, the utilization rate of a certain role is significantly higher than that of other roles, the system will appropriately increase the resource balancing weight w 2 , thereby optimizing the reallocation of tasks and reducing the pressure on high-load roles.

[0073] When the completion rate R p (t) of high-priority tasks is lower than the set target value, the system will increase the weight w 3 , giving priority to ensuring the allocation of high-priority tasks and improving the processing speed of priority tasks.

[0074] As a possible implementation method, the logic of feedback adjustment can be quantified by the following formula: w k (t + 1) = w k (t) + γ k ·e k (t) where w k (t + 1) represents the optimized weight at time t + 1; the optimized weight w k (t) at time t, e k (t) represents the error between the monitoring index and the target value, such as Q(t) - Q threshold ; γ k is the feedback gain parameter used to control the adjustment speed.

[0075] In this embodiment, the feedback adjustment mechanism is implemented through the following steps: First, the system calculates key indicators based on the monitoring data, including Q(t), U j (t) and R p (t). These data are generated in real time from the execution status of the approval tasks and the resource usage of the approval roles.

[0076] Second, according to the difference between the calculation result and the preset target, the system triggers the corresponding feedback rules. For example, when it is detected that Q(t) continues to increase, the system will gradually increase the value of w 1 , while decreasing the values of w 2 and w 3 .

[0077] Finally, after updating and optimizing the weights, the system runs the optimization algorithm again to adjust the task allocation scheme. For example, by updating the task allocation matrix X, the task t is reallocated i to the low-load role r j .

[0078] For step S6, generally, the final performance of the credit approval pipeline needs to be quantitatively displayed through specific result data, and at the same time, the optimization effect of the model is comprehensively evaluated. By outputting the optimized task allocation scheme, key performance indicators, and model operation performance, it can provide a basis for the subsequent adjustment and improvement of the system. This step of the present invention can not only effectively verify the technical effects of the foregoing steps, but also discover potential problems through the analysis of the results to further optimize the process

[0079] In this embodiment, the result output content mainly includes the task allocation scheme, key performance indicators, and model evaluation results. Specifically The optimized task allocation scheme is output in the form of the allocation matrix X X = [x i,j , x i,j ∈ {0, 1} where x i,j = 1 indicates that the task t i is assigned to the approval role r j ; x i,j = 0 indicates that the task t i is not assigned to the role r j ; each row of the matrix X represents the allocation result of a task

[0080] As an option, the system can also output the task allocation information in tabular form. For example, each record includes information such as task identifier, assigned role, estimated processing time, priority weight, and completion time

[0081] The output of the key performance indicators includes approval efficiency, resource utilization rate, and high-priority task completion rate. The specific calculation formulas are as follows The approval efficiency is represented by the total completion time T total : The resource utilization rate is represented by the average utilization rate of each role : The high-priority task completion rate R p is expressed as where n is the total number of tasks, m is the total number of approval roles, Ci,j Denote task t i The processing time x on role r j Allocate variable U for the task i,j Denote the utilization rate of role r, Capacity(r j Denote role r j The maximum resource processing capacity of the approval role r, T j Denote the high-priority task set j Denote the task completion status, |T p | is the total number of high-priority tasks Denote the task completion status, |T p | is the total number of high-priority tasks

[0082] The model evaluation results focus on the running efficiency and optimization effect, and output the convergence time of the optimized model, the number of population iterations, and the number of Pareto optimal solutions, etc.

[0083] In this embodiment, the system evaluation conducts a quantitative analysis of the optimization effect and running performance of the model, including the following specific contents In terms of the optimization effect, mainly analyze the achievement of the optimization goal. For example, by comparing the approval efficiency T total , resource utilization rate U j and the completion rate R of priority tasks p , evaluate the improvement amplitude of the model

[0084] In terms of the running performance, evaluate the running efficiency of the optimization algorithm. For example, by counting the number of population iterations, convergence time, and population diversity of the non-dominated sorting genetic algorithm, verify the robustness and adaptability of the algorithm

[0085] As a possible implementation, the system evaluation can also perform stress tests in combination with dynamic change scenarios. For example, in scenarios simulating high task loads or uneven resource distributions, test the performance of the model under different weight configurations

[0086] In this embodiment, the result analysis includes the following aspects Specifically, the output of the task allocation scheme can be compared with the original scheme to analyze the balance of the optimized task allocation. For example, count the number of tasks and processing time of each approval role to determine whether there are obvious imbalances in the allocation

[0087] For key performance indicators, line charts or bar charts can be generated through data visualization tools. For example, show the change trend of the approval efficiency T total with the number of task arrivals, or compare the differences in the resource utilization rate U j under different weight configurations

[0088] As an extended function, the result analysis can be further refined to specific task dimensions. For example, by analyzing the completion time distribution of high-priority tasks, it can be determined whether the response speed of high-priority tasks meets expectations.

[0089] The output form of the results can be a report or a visualization chart. For example: The report includes the task assignment results, summary of performance metrics, and model evaluation data; The visualization chart includes a heat map of task assignment, a bar chart of resource utilization, and a line chart of approval efficiency.

[0090] A pipeline design system for small business credit based on the credit factory model described below can be correspondingly referred to in relation to a distributed management method for a cloud container cluster described above.

[0091] Please refer to the appendix Figure 2 The present invention also provides a pipeline design system for small business credit based on the credit factory model. By combining a multi-objective optimization model and a dynamic game model, and through a modular division of labor and real-time data processing mechanism, the system can efficiently manage the dynamic allocation and optimization of credit approval tasks, achieving improved approval efficiency, balanced resource utilization, and guarantee of high-priority tasks.

[0092] A data collection module for obtaining basic data of a credit approval task set and an approval role set; Collecting attribute information such as the processing time, priority weight, and latest completion time of tasks; Collecting resource capabilities and current load information of approval roles; In one embodiment, the data collection module supports dynamically synchronizing information from external systems (such as a credit assessment platform or a credit information data center) to achieve real-time update of task and role data.

[0093] An optimization calculation module for establishing an optimization model and generating an initial task assignment plan; According to the multi-objective optimization method, an optimization model is constructed to comprehensively consider the approval time, resource utilization rate, and completion rate of priority tasks; in one possible implementation, this module can flexibly optimize task assignment by combining a weight adjustment mechanism. Specifically, this module supports generating a set of Pareto optimal solutions through a non-dominated sorting genetic algorithm and selecting the optimal task assignment plan as the initial result.

[0094] A dynamic scheduling module for adjusting the task assignment plan according to the dynamic changes of tasks; When a new task arrives or the status of an existing task changes, the dynamic scheduling module recalculates the task assignment plan; The module calculates the revenue function of the approval role based on the dynamic game model and realizes the optimal allocation of tasks by solving the Nash equilibrium; In one embodiment, the module supports the task reallocation function to adapt to the approval peak or uneven utilization of role resources.

[0095] The real-time monitoring module is used to monitor the running status of the approval pipeline and provide a feedback basis for task allocation; The monitored metrics include the approval queue length, role resource utilization rate, and high-priority task completion rate; The module can perform real-time analysis on the monitored data, discover potential bottlenecks, and trigger the weight adjustment mechanism; In a possible implementation, the real-time monitoring module can display the monitoring results in the form of charts to provide an intuitive feedback on the system running status.

[0096] The data output module is used to output the task allocation results and system performance metrics; Output the task allocation matrix, which contains the specific allocation relationship between tasks and approval roles; Output key performance metrics, such as the total approval time, resource utilization distribution, and completion of priority tasks; In one embodiment, the data output module supports generating a report document, including the comparison of allocation schemes, the trend of metric changes, and the analysis of system optimization effects.

[0097] In this implementation, through the division of labor and cooperation of each module, this system realizes the efficient optimization and dynamic adjustment of the credit approval pipeline: The data acquisition module ensures the integrity and real-time nature of the basic data and provides accurate input for the optimization model; The optimization calculation module lays the foundation for task allocation by generating the initial task allocation scheme; The dynamic scheduling module adapts to the dynamic changes of tasks and improves the flexibility of the system; The real-time monitoring module realizes the visualization of the running status and the active warning of bottleneck problems through the acquisition and feedback of status data; the data output module intuitively displays the optimization effect of the system and provides a basis for further optimization.

[0098] In one embodiment, the system is deployed in the multi-branch approval scenario to realize the unified allocation of tasks across branches and resource optimization. In another embodiment, the system is used for the credit approval pipeline of small and micro enterprises to improve the response speed of high-priority tasks through dynamic scheduling and balance the resource utilization rate at the same time.

[0099] The system of this embodiment can be used to execute the above method embodiments, and its principle and technical effects are similar, so they will not be elaborated here.

[0100] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A production line design method based on the small business credit factory model, characterized in that: The following steps are involved: Initialize task and resource parameters, collect credit approval task sets and approval role sets, define processing time, priority weight and latest completion time for each task, and define resource capacity and initial load for each approval role; Construct a multi-objective optimization model, comprehensively consider task completion time, resource utilization and task priority completion quality, and transform the optimization objectives into a comprehensive objective function with weights; Use optimization algorithms to solve the comprehensive objective function, generate task allocation plans, and complete initial task allocation based on constraints; Dynamically adjust the task allocation plan. When a new task arrives, calculate the benefit function of each approval role by building a dynamic game model to optimize and solve its task allocation strategy; Monitor the approval queue length, resource utilization, and task completion time in real time, dynamically adjust optimization weights based on monitoring results, and further optimize task allocation plans; Output the final task allocation results, including approval efficiency, resource utilization and priority completion status.

2. According to claim 1, a pipeline design method based on the small business credit factory model is characterized in that: The optimization objectives include: Minimize the total completion time of the credit approval pipeline by calculating the time spent on all tasks among different approval roles; maximize the resource utilization of approval roles by increasing the ratio of the task load assigned to approval roles to their resource capacity; maximize the quality of task priority completion by giving priority to higher priority tasks and increasing the completion ratio of high priority tasks.

3. According to claim 1, a pipeline design method based on the small business credit factory model is characterized in that: The multi-objective optimization model includes the following constraints: Each task must be assigned to one and only one approval role; The total load of the approval role must not exceed its resource capacity; The completion time of each task must not exceed its latest completion time.

4. According to claim 1, a pipeline design method based on the small business credit factory model is characterized in that: The comprehensive objective function is solved by using a non-dominated sorting genetic algorithm, comprising the following steps: Initialize the population and generate multiple individuals according to the random task allocation scheme. Each individual is a task allocation matrix X = [x i,j ], where x i,j ∈{0,1} represents task t i Is it assigned to the approval role? j ; Based on the comprehensive objective function F, each individual in the population is non-dominated and sorted. The comprehensive objective function is defined as: Among them, C i,j For task t i In the approval role j Processing time; Capacity(r j ): Approval role j Resource capability; P i :Task t i The priority weight; w1, w2, w3 are weight coefficients; C i,j For task t i In the approval role j Estimated processing time on; x i,j Represents task t i Is it assigned to the approval role? j ; n is the total number of tasks; m is the total number of approval roles; Capacity(r j ) is the approval role r j The maximum resource capacity, P i Task i The priority weight of Perform genetic operations, including selection, crossover, and mutation, to generate a new generation of population: Selection operation: select individuals by probability according to fitness value; Crossover operation: randomly exchange the task allocation matrix of the selected individuals to generate a new task allocation scheme; Mutation operation: Randomly adjust the x in the task allocation matrix i,j to explore new solution spaces; Iterate repeatedly until the population converges and obtains the Pareto optimal solution set; Select the solution that suits the current constraints from the Pareto optimal solution set as the initial task allocation solution.

5. The assembly line design method based on the small business credit factory model according to claim 1 is characterized in that: The task allocation scheme includes: Based on the dynamic game model, the approval role is regarded as a game participant, and the profit function is constructed through task priority completion quality and resource utilization; By optimizing and solving the Nash equilibrium of the game model, the optimal task allocation strategy for each approval role is determined.

6. The assembly line design method based on the small business credit factory model according to claim 1 is characterized in that: The function of the dynamic game model is expressed by the following formula: Approval Role j The profit function U j It consists of two parts: task priority completion quality and resource utilization, and is defined as: Among them, T j Assigned to role r j The task set P i For task t i Priority weight; Capacity(r j ) is the approval role r j Resource capacity; L j For the approval role j The current load; α, β are weight parameters used to balance the impact of task priority completion quality and resource utilization; The goal of the game model is to make the benefits of all approval roles reach Nash equilibrium, that is, the benefits of each role are maximized when the strategies of other roles are fixed, and the following conditions are met: in, Denotes the profit function U j For the role j The partial derivative of the strategy, R is the set of approval roles; The Nash equilibrium solution is achieved through the following steps: Initialize the task allocation scheme T j ; For each character r j , optimize the profit function U while fixing the strategies of other roles j ; The allocation scheme is updated iteratively until the payoff functions of all roles converge to an equilibrium state.

7. The assembly line design method based on the small business credit factory model according to claim 1 is characterized in that: The real-time monitoring includes: Monitor the length of the approval queue to determine the backlog of tasks; Monitor resource utilization of approval roles to optimize task allocation; Adjust and optimize weights based on monitoring results to cope with approval peaks.

8. The assembly line design method based on the small business credit factory model according to claim 1 is characterized in that: The adjustment of the optimization weights satisfies the following conditions: When the approval queue length exceeds the preset threshold, the weight of approval efficiency is increased; When resource utilization is uneven, increase the weight of resource utilization; When the completion rate of priority tasks is low, increase the weight of the completion quality of priority tasks.

9. The assembly line design method based on the small business credit factory model according to claim 1 is characterized in that: The allocation results include: The final task allocation plan includes the approval role allocation information for each task; Key performance indicators of the credit approval pipeline, including total approval time, resource utilization, and priority task completion.

10. The assembly line design system based on the small business credit factory mode is applied to the assembly line design method based on the small business credit factory mode according to any one of claims 1 to 9, characterized in that: include: Data collection module, used to collect relevant data of credit approval task set and approval role set; Optimization calculation module, used to build optimization model and solve comprehensive objective function; Dynamic scheduling module, used to adjust task allocation scheme based on dynamic game model; Monitoring and feedback module, used to monitor the approval queue, resource utilization and task completion in real time, and dynamically adjust the optimization weight; Data output module, used to output the final task allocation results and key performance indicators.

Images