Micro-service and ai-based financing lease supply chain management optimization method

Abstract

The application relates to the field of supply chain management, in particular to a financing lease supply chain management optimization method based on microservices and AI, which comprises a supply chain simulation module, a demand planning module, a microservice module, an optimization deployment module and a balanced scheduling module, the supply chain simulation module is used for constructing a supply chain model, the demand planning module is used for initializing the positions of leased goods, the microservice module is used for constructing a microservice cluster, the optimization deployment module is used for adjusting the flow of node goods, and the balanced scheduling module is used for selecting a scheduling strategy and balancing node loads, the application can help a lease platform to manage supply chain information, improve equipment management and operation level, enhance supply chain cooperation, improve the robustness and working efficiency of microservice nodes, improve the efficiency of processing supply-demand uncertainty and network flow optimization problems in a lease process, improve the decision-making efficiency and accuracy of supply chain scheduling management, and improve the efficiency of financing lease.

Description

Technical Field

[0001] This invention relates to the field of supply chain management, specifically to a method for optimizing supply chain management in financial leasing based on microservices and AI. Background Technology

[0002] Financial leasing is a type of leasing that transfers ownership of assets. When a lessee is unable to purchase equipment, they can apply to a financial leasing platform to purchase the equipment and lease it out. The lessee acquires the equipment by paying rent on schedule, using the revenue generated by the equipment to pay the rent until the usage fees cover the platform's purchase cost, thus achieving the transfer of ownership. The advantage of financial leasing is that it allows lessees to avoid ownership risks associated with the equipment, while simultaneously reducing the platform's financing risks. Microservices are a discrete development architecture that breaks down complex applications into smaller, independently running service programs that collaborate to complete complex tasks.

[0003] In the process of developing a financial leasing platform, since the leased assets are uniformly managed by the platform, the platform needs to deal with a large number of lease applications, high-frequency flow of leased assets, and real-time supply and allocation tasks. Ordinary service processing algorithms are difficult to flexibly schedule leased assets. Processing through microservice algorithms can speed up the scheduling speed, but the compatibility between microservice nodes and the leasing supply chain is insufficient, precise deployment is difficult, and the processing capacity of a single microservice node is limited, making it difficult for high-traffic platforms to use.

[0004] In addition, the operation of the leasing supply chain is easily affected by environmental changes or transportation processes, resulting in uncontrollable losses. Issues such as lease suspension, buyout, and cross-regional leasing may also occur during the supply process, causing uncertain demand factors to affect the normal operation of the supply chain and increasing the difficulty of planning and allocating leased assets. Summary of the Invention

[0005] The purpose of this invention is to provide a microservices and AI-based optimization method for financial leasing supply chain management to solve the problems mentioned in the background.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a financial leasing supply chain management optimization system based on microservices and AI, comprising: a supply chain simulation module, a demand planning module, a microservice module, an optimization deployment module, and a load balancing scheduling module; The supply chain simulation module is used to input leasing information into the database, simulate the flow of leased goods and changes in demand using discrete event simulation, and identify controllable factors in the leasing process through hypercube sampling evaluation, including lease payment terms, order batches, transportation routes and inventory allocation. The leasing center is modeled as a node, the flow of leased goods is a directed edge, and the controllable factors constitute decision variables. Based on leasing costs and leasing time, a total cost function is defined, and a model regularization term is added through the R-Drop algorithm to adjust the scenario index and construct the supply chain model. The demand planning module is used to estimate the demand distribution of each leasing node in the supply chain based on the changing state of leasing demand, construct a set with uncertain demand, transform the uncertain demand into a deterministic form based on Fenchel duality, transform the robust optimization constraint into a second-order cone constraint, transform the global supply chain model into a second-order cone programming model, and use the SOCP solver to solve the transformed model to obtain the supply chain local planning model. The microservice module is used to analyze the leasing service process through the KANO model, decompose the leasing service process into sub-processes, map each part of the supply chain model to the corresponding sub-processes, discretize each sub-process, build a microservice cluster, and merge and split the microservices in real time by quantifying the cohesion and coupling between microservices. The corrected microservices are then encapsulated into independent control nodes, which take input request information as input and decision variables as output. The optimized deployment module is used to collect request data for each microservice, use a seasonal ARIMA model to perform time-series prediction on the time-series request data of each microservice, estimate the request arrival rate of each microservice in the next planning period, calculate the cost and expected duration of the microservice on the node based on the liquidity of node leased resources, determine the microservice deployment plan with the goal of minimizing the cost of supplying leased items at the microservice location and the constraint that the expected duration is below a threshold, and adjust the flow of node resources by controlling decision variables. The balanced scheduling module is used to input leasing service requests into microservice programs, construct microservice call graphs, model the dependencies between microservices using multi-hop graph attention networks, input the time series data of each microservice program into gated loop units to model the time characteristics of microservices, divide container groups according to time characteristics using spectral clustering algorithms, determine the state space, action space and reward function within each container group using an AI deep learning network based on the DQN algorithm, plan a scheduling scheme with the goal of maximizing the reward function, select the optimal scheduling strategy for resource scheduling, balance node load, and automatically output leased item supply information.

[0007] Furthermore, the supply chain simulation module includes: a leasing information unit and a graph simulation unit; The rental information unit is used to build a relational database and input rental information, which includes: rental center location, type of leased item, rental cost, and rental period; The graph simulation unit is used to simulate leasing demand, identify controllable variables in the supply process, and construct a supply chain model with leasing centers with inventory capacity and order processing capabilities as nodes and the flow of leased goods as a dependency.

[0008] Furthermore, the demand planning module includes: an uncertainty handling unit, a model transformation unit, and a local modeling unit; The uncertainty processing unit is used to describe the uncertainties in the leasing process based on historical data, including: lease termination, buyout, lease renewal and material transfer, and to establish a set of uncertain demands; The model transformation unit is used to replace the random demand in the supply chain model with an uncertain set to obtain robust optimization constraints. It uses Fenchel duality to define the conjugate model and transforms the conjugate model through auxiliary variables and dual variables to obtain a second-order cone constraint model. The local modeling unit is used to input the second-order cone constraint model into the SOCP solver, output the node local decisions, including inventory level and replenishment trigger point, initialize the leased item location according to the node local decisions, and establish the supply chain local planning model.

[0009] Furthermore, the microservice module includes: a service discrete unit and a quantitative coupling unit; The service discretization unit is used to classify users’ needs for leasing services through the KANO model to obtain discretized processing procedures, including: leasing request, leasing item retrieval, leasing equipment purchase, rent payment, ownership purchase and leasing termination request. In each processing procedure, secondary discretization is performed according to the computing power requirements of the service, and the service is encapsulated into a microservice program with fixed computing power. The quantitative coupling unit is used to label nodes with microservice processing capabilities and their corresponding local planning models, thus mapping the microservice cluster.

[0010] Furthermore, the optimized deployment module includes: a solution expansion unit and an integrated deployment unit; The solution extension unit is used to predict future request arrival rates through the API gateway logs of microservices, and calculate the cost and expected duration required to satisfy the service based on the lease inventory of each node and the resource liquidity between each node. The integrated deployment unit is used to distribute the microservices across the supply chain nodes, ensuring that the expected execution time of any microservice is lower than a threshold, and selecting the deployment scheme with the lowest supply cost.

[0011] Furthermore, the balanced scheduling module includes: an AI training unit, a gated loop unit, and a scheduling processing unit; The AI ​​training unit is used to record every run of the microservice, construct a call graph of all microservices within the cycle, and use a graph attention network to learn the multi-hop dependencies between microservices to obtain the embedded representation of each microservice. The gated loop unit is used to input the time-series data of the microservices into the gated loop unit. The time-series data includes request arrival rate and resource utilization rate, and outputs the usage frequency of each microservice. The scheduling processing unit is used to perform spectral clustering on microservices, divide microservices with the same dependency into container groups, design a scheduling AI within each container group, and output the state space, action space and reward function within each container group. The reward function is a weighted sum of microservice call time, leased item call cost and total lease revenue, and plans the optimal scheduling scheme.

[0012] A microservices- and AI-based approach to optimizing supply chain management in the financial leasing industry includes the following steps: Step S1. Enter the leasing information into the database, simulate leasing demand, identify controllable variables in the supply process, and build a supply chain model with leasing centers with inventory capacity and order processing capabilities as nodes and the flow of leased goods as the dependency relationship. Step S2. Based on historical data, establish an uncertain demand set, replace the random demand in the supply chain model with the uncertain set, and obtain a second-order cone constraint model through Fenchel transformation. Input the second-order cone constraint model into the SOCP solver, output the node local decision, and initialize the leased item location according to the local decision. Step S3. Decompose the rental service process into sub-processes, discretize each sub-process, and encapsulate it into a microservice program with fixed computing power. At the same time, according to the cohesion and coupling between the quantified microservices, merge and split the program in real time. Step S4. Predict the request arrival rate through microservice logs, calculate the cost and expected duration required to satisfy the service based on the node lease inventory and resource liquidity, select the solution with the lowest supply cost and the time requirement, and distribute each microservice in the supply chain nodes. Step S5. Based on the microservice running status, learn the multi-hop dependencies between microservices, perform spectral clustering on the microservices, divide the container groups according to the clustering results, and output the state space, action space and reward function of each container group through scheduling AI based on the usage frequency of each microservice. Plan the leasing scheduling scheme with the highest reward function as the goal.

[0013] Furthermore, step S1 includes: Step S11. Construct a relational database and enter the rental information, which includes: rental center location, type of leased property, rental cost, and rental period; Step S12. Use discrete event simulation to simulate the flow of leased assets and changes in demand, and use hypercube sampling to evaluate and identify controllable factors in the leasing process, including lease payment terms, order batches, transportation routes and inventory allocation; Step S13. Model the leasing center as a node, the flow of leased goods as directed edges, and controllable factors as decision variables. Based on leasing costs and leasing time, define the total cost function, add model regularization terms using the R-Drop algorithm, adjust the scenario index, and construct the supply chain model.

[0014] Furthermore, step S2 includes: Step S21. Describe the uncertainties in the leasing process based on historical data, including: lease termination, buyout, lease renewal and material relocation, establish an uncertain demand set, replace the random demand in the supply chain model with the uncertain set, and obtain robust optimization constraints; Step S22. Use Fenchel dual definition to define the conjugate model. Transform the conjugate model through auxiliary variables and dual variables to obtain a second-order cone constraint model. Input the second-order cone constraint model into the SOCP solver and output the node local decisions, including inventory level and replenishment trigger point. Initialize the leased item location according to the node local decisions and establish the supply chain local planning model.

[0015] Furthermore, step S3 includes: Step S31. Classify the user's demand for leasing services using the KANO model to obtain a discretized processing procedure, including: leasing request, leasing item retrieval, leasing equipment purchase, rent payment, ownership purchase and lease termination request. In each processing procedure, perform secondary discretization according to the computing power requirements of the service, and encapsulate the service into a microservice program with fixed computing power. Step S32. Map each part of the supply chain model to the corresponding sub-process. By quantifying the cohesion and coupling between microservices, merge and split the microservices in real time. Encapsulate the corrected microservices into independent control nodes. The control nodes take input request information as input and decision variables as output.

[0016] Furthermore, step S4 includes: Step S41. Collect request data for each microservice, use a seasonal ARIMA model to perform time-series prediction on the time-series request data of each microservice, and estimate the request arrival rate of each microservice in the next planning period. Step S42. Based on the liquidity of node leased resources, calculate the cost and expected duration of the microservice on the node. With the goal of minimizing the cost of supplying leased assets at the microservice location and the constraint that the expected duration is below a threshold, determine the microservice deployment plan.

[0017] Furthermore, step S5 includes: Step S51. Input the rental service request into the microservice program, construct the microservice call graph, use a multi-hop graph attention network to model the dependencies between microservices, input the time series data of each microservice program into the gated loop unit, and model the time characteristics of the microservice. Step S52. Divide the container groups according to time characteristics using the spectral clustering algorithm, and determine the state space, action space and reward function in each container group using the AI ​​deep learning network based on the DQN algorithm. The reward function is a weighted sum of microservice call time, leased item call cost and total lease revenue. Step S53. Plan a scheduling scheme with the goal of maximizing the reward function, select the optimal scheduling strategy for resource scheduling, balance the node load, and automatically output the supply information of leased items.

[0018] Compared with the prior art, the beneficial effects achieved by the present invention are: 1. This invention uses simulation analysis to screen controllable factors, construct a supply chain model, and build a set of uncertain demands based on the changing state of leasing demand. Based on cone duality theory, the global model is transformed into a second-order cone programming model, which can help leasing platforms manage leased asset information and supply chain information, effectively improve equipment management and operation level, increase user leasing satisfaction, enhance supply chain collaboration, and improve the profitability and sustainability of financial leasing.

[0019] 2. This invention can discretize leasing services, filter out leasing service processes, and correct the microservice partitioning results based on quantitative indicators of microservice coupling. It constructs microservice nodes that handle different services and deploys them in the supply chain model, realizing an automated process from demand acquisition to optimized deployment, improving the robustness and efficiency of microservice nodes, and increasing the efficiency of handling supply and demand uncertainties and network flow optimization problems in the leasing process.

[0020] 3. This invention utilizes multi-hop graph attention networks to model dependencies, gated cyclic units to model time characteristics, spectral clustering algorithms to divide container groups for node resource scheduling, selects the optimal scheduling strategy, and automatically outputs leased goods supply information to achieve load balancing, thereby improving the decision-making efficiency and accuracy of supply chain scheduling management and realizing intelligent load balancing and automated resource scheduling. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the structure of the financial leasing supply chain management optimization system based on microservices and AI of this invention; Figure 2 This is a schematic diagram illustrating the steps of the financial leasing supply chain management optimization method based on microservices and AI of this invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figures 1 to 2 The present invention provides a technical solution: a financial leasing supply chain management optimization system based on microservices and AI, including: a supply chain simulation module, a demand planning module, a microservice module, an optimization deployment module, and a balance scheduling module; The supply chain simulation module is used to input leasing information into the database, simulate the flow of leased goods and changes in demand using discrete event simulation, and identify controllable factors in the leasing process through hypercube sampling evaluation, including lease payment terms, order batches, transportation routes and inventory allocation. The leasing center is modeled as a node, the flow of leased goods is a directed edge, and the controllable factors constitute decision variables. Based on leasing costs and leasing time, a total cost function is defined, and a model regularization term is added through the R-Drop algorithm to adjust the scenario index and construct the supply chain model. The supply chain simulation module includes: a leasing information unit and a graph simulation unit; The rental information unit is used to build a relational database and input rental information, which includes: rental center location, type of leased item, rental cost, and rental period; The graph simulation unit is used to simulate leasing demand, identify controllable variables in the supply process, and construct a supply chain model with leasing centers with inventory capacity and order processing capabilities as nodes and the flow of leased goods as a dependency.

[0024] The demand planning module is used to estimate the demand distribution of each leasing node in the supply chain based on the changing state of leasing demand, construct a set with uncertain demand, transform the uncertain demand into a deterministic form based on Fenchel duality, transform the robust optimization constraint into a second-order cone constraint, transform the global supply chain model into a second-order cone programming model, and use the SOCP solver to solve the transformed model to obtain the supply chain local planning model. The demand planning module includes: an uncertainty handling unit, a model transformation unit, and a local modeling unit; The uncertainty processing unit is used to describe the uncertainties in the leasing process based on historical data, including: lease termination, buyout, lease renewal and material transfer, and to establish a set of uncertain demands; The model transformation unit is used to replace the random demand in the supply chain model with an uncertain set to obtain robust optimization constraints. It uses Fenchel duality to define the conjugate model and transforms the conjugate model through auxiliary variables and dual variables to obtain a second-order cone constraint model. The local modeling unit is used to input the second-order cone constraint model into the SOCP solver, output the node local decisions, including inventory level and replenishment trigger point, initialize the leased item location according to the node local decisions, and establish the supply chain local planning model.

[0025] The microservice module is used to analyze the leasing service process through the KANO model, decompose the leasing service process into sub-processes, map each part of the supply chain model to the corresponding sub-processes, discretize each sub-process, build a microservice cluster, and merge and split the microservices in real time by quantifying the cohesion and coupling between microservices. The corrected microservices are then encapsulated into independent control nodes, which take input request information as input and decision variables as output. The microservice module includes: a service discrete unit and a quantitative coupling unit; The service discretization unit is used to classify users’ needs for leasing services through the KANO model to obtain discretized processing procedures, including: leasing request, leasing item retrieval, leasing equipment purchase, rent payment, ownership purchase and leasing termination request. In each processing procedure, secondary discretization is performed according to the computing power requirements of the service, and the service is encapsulated into a microservice program with fixed computing power. The quantitative coupling unit is used to label nodes with microservice processing capabilities and their corresponding local planning models, thus mapping the microservice cluster.

[0026] The optimized deployment module is used to collect request data for each microservice, use a seasonal ARIMA model to perform time-series prediction on the time-series request data of each microservice, estimate the request arrival rate of each microservice in the next planning period, calculate the cost and expected duration of the microservice on the node based on the liquidity of node leased resources, determine the microservice deployment plan with the goal of minimizing the cost of supplying leased items at the microservice location and the constraint that the expected duration is below a threshold, and adjust the flow of node resources by controlling decision variables. The optimized deployment module includes: a solution expansion unit and an integrated deployment unit; The solution extension unit is used to predict future request arrival rates through the API gateway logs of microservices, and calculate the cost and expected duration required to satisfy the service based on the lease inventory of each node and the resource liquidity between each node. The integrated deployment unit is used to distribute the microservices across the supply chain nodes, ensuring that the expected execution time of any microservice is lower than a threshold, and selecting the deployment scheme with the lowest supply cost.

[0027] The balanced scheduling module is used to input leasing service requests into microservice programs, construct microservice call graphs, model the dependencies between microservices using multi-hop graph attention networks, input the time series data of each microservice program into gated loop units to model the time characteristics of microservices, divide container groups according to time characteristics using spectral clustering algorithms, determine the state space, action space and reward function within each container group using an AI deep learning network based on the DQN algorithm, plan a scheduling scheme with the goal of maximizing the reward function, select the optimal scheduling strategy for resource scheduling, balance node load, and automatically output leased item supply information.

[0028] The balanced scheduling module includes: an AI training unit, a gated loop unit, and a scheduling processing unit; The AI ​​training unit is used to record every run of the microservice, construct a call graph of all microservices within the cycle, and use a graph attention network to learn the multi-hop dependencies between microservices to obtain the embedded representation of each microservice. The gated loop unit is used to input the time-series data of the microservices into the gated loop unit. The time-series data includes request arrival rate and resource utilization rate, and outputs the usage frequency of each microservice. The scheduling processing unit is used to perform spectral clustering on microservices, divide microservices with the same dependency into container groups, design a scheduling AI within each container group, and output the state space, action space and reward function within each container group. The reward function is a weighted sum of microservice call time, leased item call cost and total lease revenue, and plans the optimal scheduling scheme.

[0029] A microservices- and AI-based approach to optimizing supply chain management in the financial leasing industry includes the following steps: Step S1. Enter the leasing information into the database, simulate leasing demand, identify controllable variables in the supply process, and build a supply chain model with leasing centers with inventory capacity and order processing capabilities as nodes and the flow of leased goods as the dependency relationship. Step S1 includes: Step S11. Construct a relational database and enter the rental information, which includes: rental center location, type of leased property, rental cost, and rental period; Step S12. Use discrete event simulation to simulate the flow of leased assets and changes in demand, and use hypercube sampling to evaluate and identify controllable factors in the leasing process, including lease payment terms, order batches, transportation routes and inventory allocation; Step S13. Model the leasing center as a node, the flow of leased goods as directed edges, and controllable factors as decision variables. Based on leasing costs and leasing time, define the total cost function, add model regularization terms using the R-Drop algorithm, adjust the scenario index, and construct the supply chain model.

[0030] Step S2. Based on historical data, establish an uncertain demand set, replace the random demand in the supply chain model with the uncertain set, and obtain a second-order cone constraint model through Fenchel transformation. Input the second-order cone constraint model into the SOCP solver, output the node local decision, and initialize the leased item location according to the local decision. Step S2 includes: Step S21. Describe the uncertainties in the leasing process based on historical data, including: lease termination, buyout, lease renewal and material relocation, establish an uncertain demand set, replace the random demand in the supply chain model with the uncertain set, and obtain robust optimization constraints; Step S22. Use Fenchel dual definition to define the conjugate model. Transform the conjugate model through auxiliary variables and dual variables to obtain a second-order cone constraint model. Input the second-order cone constraint model into the SOCP solver and output the node local decisions, including inventory level and replenishment trigger point. Initialize the leased item location according to the node local decisions and establish the supply chain local planning model.

[0031] Step S3. Decompose the rental service process into sub-processes, discretize each sub-process, and encapsulate it into a microservice program with fixed computing power. At the same time, according to the cohesion and coupling between the quantified microservices, merge and split the program in real time. Step S3 includes: Step S31. Classify the user's demand for leasing services using the KANO model to obtain a discretized processing procedure, including: leasing request, leasing item retrieval, leasing equipment purchase, rent payment, ownership purchase and lease termination request. In each processing procedure, perform secondary discretization according to the computing power requirements of the service, and encapsulate the service into a microservice program with fixed computing power. Step S32. Map each part of the supply chain model to the corresponding sub-process. By quantifying the cohesion and coupling between microservices, merge and split the microservices in real time. Encapsulate the corrected microservices into independent control nodes. The control nodes take input request information as input and decision variables as output.

[0032] Step S4. Predict the request arrival rate through microservice logs, calculate the cost and expected duration required to satisfy the service based on the node lease inventory and resource liquidity, select the solution with the lowest supply cost and the time requirement, and distribute each microservice in the supply chain nodes. Step S4 includes: Step S41. Collect request data for each microservice, use a seasonal ARIMA model to perform time-series prediction on the time-series request data of each microservice, and estimate the request arrival rate of each microservice in the next planning period. Step S42. Based on the liquidity of node leased resources, calculate the cost and expected duration of the microservice on the node. With the goal of minimizing the cost of supplying leased assets at the microservice location and the constraint that the expected duration is below a threshold, determine the microservice deployment plan.

[0033] Step S5. Based on the microservice running status, learn the multi-hop dependencies between microservices, perform spectral clustering on the microservices, divide the container groups according to the clustering results, and output the state space, action space and reward function of each container group through scheduling AI based on the usage frequency of each microservice. Plan the leasing scheduling scheme with the highest reward function as the goal.

[0034] Step S5 includes: Step S51. Input the rental service request into the microservice program, construct the microservice call graph, use a multi-hop graph attention network to model the dependencies between microservices, input the time series data of each microservice program into the gated loop unit, and model the time characteristics of the microservice. Step S52. Divide the container groups according to time characteristics using the spectral clustering algorithm, and determine the state space, action space and reward function in each container group using the AI ​​deep learning network based on the DQN algorithm. The reward function is a weighted sum of microservice call time, leased item call cost and total lease revenue. Step S53. Plan a scheduling scheme with the goal of maximizing the reward function, select the optimal scheduling strategy for resource scheduling, balance the node load, and automatically output the supply information of leased items.

[0035] Example: Leasing information is entered into the database, the leasing service process is analyzed using the KANO model, the leasing service is discretized, controllable factors are screened, a supply chain model is constructed, an uncertain demand set is constructed, the global model is transformed into a second-order cone programming model based on Fenchel duality, the microservice partitioning results are corrected according to the microservice coupling, the request arrival rate of the next stage is predicted, the instance of each microservice is optimized, the microservices are deployed, a microservice call graph is constructed, and resource scheduling is performed based on DQN.

[0036] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0037] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A microservices and AI-based optimization method for supply chain management in financial leasing, characterized in that, The method includes the following steps: Step S1. Enter the leasing information into the database, simulate leasing demand, identify controllable variables in the supply process, and build a supply chain model with leasing centers with inventory capacity and order processing capabilities as nodes and the flow of leased goods as the dependency relationship. Step S2. Based on historical data, establish an uncertain demand set, replace the random demand in the supply chain model with the uncertain set, and obtain a second-order cone constraint model through Fenchel transformation. Input the second-order cone constraint model into the SOCP solver, output the node local decision, and initialize the leased item location according to the local decision. Step S3. Decompose the rental service process into sub-processes, discretize each sub-process, and encapsulate it into a microservice program with fixed computing power. At the same time, according to the cohesion and coupling between the quantified microservices, merge and split the program in real time. Step S4. Predict the request arrival rate through microservice logs, calculate the cost and expected duration required to satisfy the service based on the node lease inventory and resource liquidity, select the solution with the lowest supply cost and the time requirement, and distribute each microservice in the supply chain nodes. Step S5. Based on the microservice running status, learn the multi-hop dependency relationship between microservices, perform spectral clustering on microservices, divide container groups according to the clustering results, and output the state space, action space and reward function of each container group through scheduling AI according to the usage frequency of each microservice. Plan the rental scheduling scheme with the highest reward function as the goal. Step S1 includes: Step S11. Construct a relational database and enter rental information, including: rental center location, type of leased property, rental cost, and rental period; Step S12. Use discrete event simulation to simulate the flow of leased assets and changes in demand, and use hypercube sampling to evaluate and identify controllable factors in the leasing process, including lease payment terms, order batches, transportation routes and inventory allocation; Step S13. Model the leasing center as a node, the flow of leased goods as directed edges, and controllable factors as decision variables. Based on leasing cost and leasing time, define the total cost function, add model regularization terms through the R-Drop algorithm, adjust the scenario index, and build a supply chain model. Step S2 includes: Step S21. Describe the uncertainties in the leasing process based on historical data, including: lease termination, buyout, lease renewal and material relocation, establish an uncertain demand set, replace the random demand in the supply chain model with the uncertain set, and obtain robust optimization constraints; Step S22. Use Fenchel dual definition to define the conjugate model, transform the conjugate model through auxiliary variables and dual variables to obtain the second-order cone constraint model, input the second-order cone constraint model into the SOCP solver, output the node local decisions, including inventory level and replenishment trigger point, initialize the leased item location according to the node local decisions, and establish the supply chain local planning model. Step S4 includes: Step S41. Collect request data for each microservice, use a seasonal ARIMA model to perform time-series prediction on the time-series request data of each microservice, and estimate the request arrival rate of each microservice in the next planning period. Step S42. Based on the liquidity of node leased resources, calculate the cost and expected duration of the microservice on the node, with the goal of minimizing the cost of supplying leased assets at the microservice location and the constraint that the expected duration is below a threshold, to determine the microservice deployment plan; Step S5 includes: Step S51. Input the rental service request into the microservice program, construct the microservice call graph, use a multi-hop graph attention network to model the dependencies between microservices, input the time series data of each microservice program into the gated loop unit, and model the time characteristics of the microservice.

2. The method for optimizing supply chain management of financial leasing based on microservices and AI according to claim 1, characterized in that: Step S3 includes: Step S31. Classify the user's demand for leasing services using the KANO model to obtain a discretized processing procedure, including: leasing request, leasing item retrieval, leasing equipment purchase, rent payment, ownership purchase and lease termination request. In each processing procedure, perform secondary discretization according to the computing power requirements of the service, and encapsulate the service into a microservice program with fixed computing power. Step S32. Map each part of the supply chain model to the corresponding sub-process. By quantifying the cohesion and coupling between microservices, merge and split the microservices in real time. Encapsulate the corrected microservices into independent control nodes. The control nodes take input request information as input and decision variables as output.

3. The method for optimizing supply chain management of financial leasing based on microservices and AI according to claim 2, characterized in that: Step S5 includes: Step S52. Divide the container groups according to time characteristics using the spectral clustering algorithm, and determine the state space, action space and reward function in each container group using the AI ​​deep learning network based on the DQN algorithm. The reward function is a weighted sum of microservice call time, leased item call cost and total lease revenue. Step S53. Plan a scheduling scheme with the goal of maximizing the reward function, select the optimal scheduling strategy for resource scheduling, balance the node load, and automatically output the supply information of leased items.

4. A microservices and AI-based supply chain management optimization system for financial leasing, used to implement the microservices and AI-based supply chain management optimization method for financial leasing as described in claim 1, characterized in that, The system includes the following modules: supply chain simulation module, demand planning module, microservice module, optimization deployment module, and load balancing module; The supply chain simulation module is used to input leasing information into the database, simulate the flow of leased goods and changes in demand using discrete event simulation, and identify controllable factors in the leasing process through hypercube sampling evaluation, including lease payment terms, order batches, transportation routes and inventory allocation. The leasing center is modeled as a node, the flow of leased goods is a directed edge, and the controllable factors constitute decision variables. Based on leasing costs and leasing time, a total cost function is defined, and a model regularization term is added through the R-Drop algorithm to adjust the scenario index and construct the supply chain model. The demand planning module is used to estimate the demand distribution of each leasing node in the supply chain based on the changing state of leasing demand, construct a set with uncertain demand, transform the uncertain demand into a deterministic form based on Fenchel duality, transform the robust optimization constraint into a second-order cone constraint, transform the global supply chain model into a second-order cone programming model, and use the SOCP solver to solve the transformed model to obtain the supply chain local planning model. The microservice module is used to analyze the leasing service process through the KANO model, decompose the leasing service process into sub-processes, map each part of the supply chain model to the corresponding sub-processes, discretize each sub-process, build a microservice cluster, and merge and split the microservices in real time by quantifying the cohesion and coupling between microservices. The corrected microservices are then encapsulated into independent control nodes, which take input request information as input and decision variables as output. The optimized deployment module is used to collect request data for each microservice, use a seasonal ARIMA model to perform time-series prediction on the time-series request data of each microservice, estimate the request arrival rate of each microservice in the next planning period, calculate the cost and expected duration of the microservice on the node based on the liquidity of node leased resources, determine the microservice deployment plan with the goal of minimizing the cost of supplying leased items at the microservice location and the constraint that the expected duration is below a threshold, and adjust the flow of node resources by controlling decision variables. The balanced scheduling module is used to input leasing service requests into microservice programs, construct microservice call graphs, model the dependencies between microservices using multi-hop graph attention networks, input the time series data of each microservice program into gated loop units to model the time characteristics of microservices, divide container groups according to time characteristics using spectral clustering algorithms, determine the state space, action space and reward function within each container group using an AI deep learning network based on the DQN algorithm, plan a scheduling scheme with the goal of maximizing the reward function, select the optimal scheduling strategy for resource scheduling, balance node load, and automatically output leased item supply information.

5. The financial leasing supply chain management optimization system based on microservices and AI according to claim 4, characterized in that: The supply chain simulation module includes: a leasing information unit and a graph simulation unit; The rental information unit is used to build a relational database and input rental information, which includes: rental center location, type of rental item, rental cost, and rental period; The graph simulation unit is used to simulate leasing demand, identify controllable variables in the supply process, and construct a supply chain model with leasing centers with inventory capacity and order processing capabilities as nodes and the flow of leased goods as a dependency.

6. The financial leasing supply chain management optimization system based on microservices and AI according to claim 5, characterized in that: The demand planning module includes: an uncertainty handling unit, a model transformation unit, and a local modeling unit; The uncertainty processing unit is used to describe the uncertainties in the leasing process based on historical data, including: lease termination, buyout, lease renewal and material transfer, and to establish a set of uncertain demands; The model transformation unit is used to replace the random demand in the supply chain model with an uncertain set to obtain robust optimization constraints. It uses Fenchel duality to define the conjugate model and transforms the conjugate model through auxiliary variables and dual variables to obtain a second-order cone constraint model. The local modeling unit is used to input the second-order cone constraint model into the SOCP solver, output the node local decisions, including inventory level and replenishment trigger point, initialize the leased item location according to the node local decisions, and establish the supply chain local planning model.

7. The financial leasing supply chain management optimization system based on microservices and AI according to claim 6, characterized in that: The microservice module includes: a service discrete unit and a quantitative coupling unit; The service discretization unit is used to classify users’ needs for leasing services through the KANO model to obtain discretized processing procedures, including: leasing request, leasing item retrieval, leasing equipment purchase, rent payment, ownership purchase and leasing termination request. In each processing procedure, secondary discretization is performed according to the computing power requirements of the service, and the service is encapsulated into a microservice program with fixed computing power. The quantitative coupling unit is used to mark nodes with microservice processing capabilities and their corresponding local planning models, mapping microservice clusters. The optimized deployment module includes: a solution expansion unit and an integrated deployment unit; The solution extension unit is used to predict future request arrival rates through the API gateway logs of microservices, and calculate the cost and expected duration required to satisfy the service based on the lease inventory of each node and the resource liquidity between each node. The integrated deployment unit is used to distribute the microservices across the supply chain nodes, ensuring that the expected execution time of any microservice is lower than a threshold, and selecting the deployment scheme with the lowest supply cost.

8. The financial leasing supply chain management optimization system based on microservices and AI according to claim 7, characterized in that: The balanced scheduling module includes: an AI training unit, a gated loop unit, and a scheduling processing unit; The AI ​​training unit is used to record every run of the microservice, construct a call graph of all microservices within the cycle, and use a graph attention network to learn the multi-hop dependencies between microservices to obtain the embedded representation of each microservice. The gated loop unit is used to input the time-series data of the microservices into the gated loop unit. The time-series data includes request arrival rate and resource utilization rate, and outputs the usage frequency of each microservice. The scheduling processing unit is used to perform spectral clustering on microservices, divide microservices with the same dependency into container groups, design a scheduling AI within each container group, and output the state space, action space and reward function within each container group. The reward function is a weighted sum of microservice call time, leased item call cost and total lease revenue, and plans the optimal scheduling scheme.

Images