Fund routing payment system based on large model application

Abstract

The invention relates to the field of credit data management, in particular to a fund routing payment system based on large model application. Comprising an order center, and the order center is in communication connection with a data acquisition and processing module, a model construction module and a risk management module. Obtaining fund payment data of the internal and external data sources of multiple parties, and processing the fund payment data to obtain fund application data; setting a user interaction interface and a fund payment large model, and analyzing the fund application data through the fund payment large model to obtain a fund recommendation path; and performing risk assessment on the payment process in combination with the fund recommendation path to obtain a risk assessment result, managing the payment process through the risk assessment result, and interacting with the user through the user interaction interface. According to the invention, payment efficiency and security can be improved.

Description

Technical Field

[0001] This invention relates to the field of fund payment, and more specifically to a fund routing payment system based on large-scale model applications. Background Technology

[0002] With the rapid development of the digital economy, new payment scenarios such as mobile payment, cross-border payment, supply chain finance, and digital currency are constantly emerging, and payment systems are gradually evolving towards intelligence, high speed, diversification, and high reliability. Against this backdrop, intelligent optimization and security of fund flow paths have become one of the key technological challenges. At the same time, the participants in the payment market are becoming increasingly diverse, and the payment network architecture is becoming increasingly complex. Traditional static routing mechanisms are struggling to meet the real-time and flexibility requirements of high concurrency, large-scale, and multi-channel fund scheduling.

[0003] Chinese patent CN112258171B discloses a routing method, system, medium, and device for an off-chain payment center based on blockchain. The method includes: dividing each transaction in the payment channel network into a series of independently routed transaction units; the payment center selecting appropriate paths to route the divided transaction units; the payment center maintaining the payment rate of transaction units in the network to maintain the capital balance of the payment channel network; and the payment center performing congestion control on the flow of transaction units in the network. This invention addresses the problems of low transaction performance, unbalanced channel capital flow, and channel capacity limiting transaction amounts in current blockchain payment channel networks. It studies routing schemes and channel capital flow rate control technologies for payment channel centers in blockchain payment channel networks, enabling larger transactions to be executed in low-capacity channels and achieving high transaction throughput without disrupting channel balance.

[0004] This invention overcomes the difficulty of properly managing transaction routing and the imbalance of funding channels in existing technologies by providing a blockchain-based off-chain payment center routing method that achieves high transaction throughput without disrupting channel balance. The specific implementation is as follows: The payment center divides each transaction in the payment channel network into a series of independently routed transaction units; the payment center selects appropriate paths to route the divided transaction units; the payment center maintains the payment rate of transaction units in the network to maintain the funding balance of the payment channel network; and the payment center performs congestion control on the flow of transaction units in the network. This off-chain payment center routing method has the following advantages: the introduction of off-chain payment channel technology effectively alleviates the scalability challenges of blockchain; and to further improve the transaction performance of the payment channel network, a payment center can be set up in the network to manage multi-channel payment routing. This invention designs a blockchain-based off-chain payment center routing method. Using the blockchain-based off-chain payment center routing method of this invention, firstly, the payment center divides each transaction in the payment channel network into a series of independently routed transaction units; secondly, the payment center selects appropriate paths to route the divided transaction units; then, the payment center maintains the payment rate of transaction units in the network to maintain the capital balance of the payment channel network; finally, the payment center performs congestion control on the flow of transaction units in the network.

[0005] The invention has the following drawbacks in the existing technology: weak security; compliance review relies on manual or template rules, resulting in slow response and error-proneness; it provides homogeneous payment services to all users; and it cannot quickly adapt to emerging payment methods; these are the problems we need to solve. Summary of the Invention

[0006] The purpose of this invention is to address the problems existing in the background technology by proposing a fund routing and payment system based on a large model application.

[0007] The technical solution of this invention: a fund routing payment system based on large-scale model applications, including an order center, wherein the order center is communicatively connected to a data acquisition and processing module, a model building module, and a risk management module.

[0008] The data acquisition and processing module is used to acquire fund payment data from multiple internal and external data sources, and to process the fund payment data to obtain fund application data.

[0009] The model building module is used to set up the user interface and the large-scale fund payment model. Through the large-scale fund payment model, fund application data is analyzed to obtain fund recommendation paths.

[0010] The risk management module is used to assess the risks of the payment process in conjunction with the recommended funding path, obtain the risk assessment results, manage the payment process based on the risk assessment results, and interact with users through the user interface.

[0011] Preferably, the process of acquiring fund payment data from multiple internal and external data sources includes:

[0012] The multiple internal and external data sources include third-party payment platforms, merchant transaction systems, bank clearing interfaces, blockchain payment channels, user terminal behavior logs, and device fingerprint information;

[0013] The payment data includes transaction data, payment channel data, and user device data; the transaction data includes transaction request data, transaction status data, historical transaction path data, and transaction purpose; the payment channel data includes payment method and congestion status; the user device data includes user identity information, device environment information, user preference data, user behavior data, and operation logs.

[0014] Preferably, the process of processing fund payment data to obtain fund application data includes:

[0015] The fund payment data is refined, including data deduplication, data completion, data correction, and data unification; fund payment data belonging to the same payment process are grouped into one category and referred to as fund payment flow data; features are extracted from the fund payment flow data to obtain fund application data; the fund application data includes transaction feature data, payment feature data, and user feature data.

[0016] Preferably, the process of setting up the user interface is as follows:

[0017] Set up a user interaction interface, which includes a user terminal, a log recording terminal, and a notification terminal;

[0018] The user interface includes a login interface, a payment interface, a transaction interface, a user settings interface, and a history interface. The login interface is used for user login and identity verification. The payment interface is used for users to input or scan a code to select payee information and to display a preview of transaction details. The transaction interface is used to display the status of the transaction after it is initiated, show the payment path and transit node status, and can refresh transaction-related data in real time. The user settings interface is used for users to set personalized parameters such as payment method, notification method, and payment limit. The history interface is used to retrieve historical payment records and analyze historical payment paths.

[0019] The log recording terminal includes a recording and storage interface and a processing interface; the recording and storage interface is used to record and store user operation logs, transaction behavior logs, and payment operation logs; the processing interface is used to perform structured format processing and categorized archiving of the logs from the storage interface.

[0020] The notification terminal is used to send payment status reminders, abnormal warnings, confirmation information, and security notifications to users.

[0021] Preferably, the process of setting up a large-scale payment model includes:

[0022] The large-scale fund payment model is connected to the user interface and is constructed by deep learning of fund application data.

[0023] The large-scale fund payment model includes a data input module and a model application module; the data input module is used to input fund application data and continuously update the large-scale fund payment model; the model application module is used to analyze and apply the fund application data, including a path optimization model, a risk model, and a user profile model.

[0024] Preferably, the process of analyzing fund application data through a large-scale fund payment model to obtain recommended fund paths includes:

[0025] The path optimization model uses a neural network model, combined with payment channels and transaction processes, to establish payment nodes and payment edges. Payment nodes are connected via payment edges to construct a payment channel graph. Deep reinforcement learning is used to process the payment channel graph, optimizing it with the maximum success rate and minimum cost as training objectives. The risk model uses Transformer technology to analyze user transaction behavior, identify risky user behaviors, and store these as risk sample behaviors within the risk model. The user profiling model acquires user feature data from the fund application data and payment channel feature data from the path optimization model. DeepFM is used to fuse and analyze the user feature data and payment channel feature data, and combined with the risk model, data exhibiting risky behaviors are eliminated, outputting user payment preferences and recommended fund paths.

[0026] Preferably, the process of conducting a risk assessment of the payment process by combining the recommended funding path and the large-scale funding payment model, and obtaining the risk assessment results, includes:

[0027] By comparing with the fund application data within the large fund payment model, path channel data and transaction behavior data of the fund recommendation path are obtained. The path channel data includes the number of payment failures and the total number of payments at each payment node. The transaction behavior data includes the number of abnormal transactions and the total number of transactions at each payment edge. Based on the path channel data and transaction behavior data, path factors and transaction factors are obtained. Payment path weights and payment transaction weights are set. Based on the payment path weights, payment transaction weights, path factors, and transaction factors, a path evaluation value is obtained.

[0028] Set risk thresholds one, two, three, and four. When the path evaluation value is less than or equal to risk threshold one, obtain risk evaluation result one. When the path evaluation value is greater than risk threshold one, obtain risk evaluation result two. When the path evaluation value is greater than or equal to risk threshold two, obtain risk evaluation result three. When the path evaluation value is greater than or equal to risk threshold three, obtain risk evaluation result four.

[0029] Preferably, the process of managing the payment process based on risk assessment results and interacting with users through a user interface includes:

[0030] When risk assessment result one is generated, the recommended funding path corresponding to risk assessment result one will be sent to the user interface to guide the user to make payment transactions according to the recommended funding path;

[0031] When Risk Assessment Result 2, Risk Assessment Result 3, or Risk Assessment Result 4 is generated, the risk data of the recommended funding path corresponding to Risk Assessment Result 2, Risk Assessment Result 3, or Risk Assessment Result 4 will be sent to the user interface to remind the user not to make payment transactions based on the risk data.

[0032] Compared with existing technologies, the technical solution of the present invention has the following beneficial technical effects: It acquires fund payment data from multiple internal and external data sources, processes the fund payment data to obtain fund application data; it sets up a user interface and a large-scale fund payment model, analyzes the fund application data through the large-scale fund payment model, and obtains a fund recommendation path; it combines the fund recommendation path to conduct risk assessment of the payment process, obtains risk assessment results, manages the payment process based on the risk assessment results, and interacts with users through the user interface; it enhances the payment system's ability to identify risky behaviors; it reduces the incidence of various risky behaviors; it improves the compliance of payment processing; it enables a smart payment experience for users; it improves user satisfaction, user stickiness, and payment success rate; and it can quickly adapt to new payment models. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of one embodiment of the present invention. Detailed Implementation

[0034] like Figure 1 As shown, the fund routing payment system based on large-scale model application proposed in this invention includes an order center, which is communicatively connected to a data acquisition and processing module, a model building module, and a risk management module.

[0035] The data acquisition and processing module is used to acquire fund payment data from multiple internal and external data sources, and to process the fund payment data to obtain fund application data.

[0036] The model building module is used to set up the user interface and the large-scale fund payment model. Through the large-scale fund payment model, fund application data is analyzed to obtain fund recommendation paths.

[0037] The risk management module is used to assess the risks of the payment process in conjunction with the recommended funding path, obtain the risk assessment results, manage the payment process based on the risk assessment results, and interact with users through the user interface.

[0038] It should be further explained that, in the specific implementation process, the process of acquiring fund payment data from multiple internal and external data sources and processing the fund payment data to obtain fund application data is as follows:

[0039] The various internal and external data sources include, but are not limited to, third-party payment platforms, merchant transaction systems, bank clearing interfaces, blockchain payment channels, user terminal behavior logs, and device fingerprint information;

[0040] The payment data includes transaction data, payment channel data, and user device data; the transaction data includes transaction request data, transaction status data, historical transaction path data, and transaction purpose; the payment channel data includes payment method and congestion status; the user device data includes user identity information, device environment information, user preference data, user behavior data, and operation logs.

[0041] The fund payment data is refined, including data deduplication, data completion, data correction, and data unification; fund payment data belonging to the same payment process are grouped into one category and referred to as fund payment flow data; features are extracted from the fund payment flow data to obtain fund application data; the fund application data includes transaction feature data, payment feature data, and user feature data.

[0042] It should be further explained that, in the specific implementation process, the process of setting up a user interface and a large-scale fund payment model, and then analyzing fund application data through the large-scale fund payment model to obtain the recommended fund path is as follows:

[0043] Set up a user interaction interface, which includes a user terminal, a log recording terminal, and a notification terminal;

[0044] The user interface includes a login interface, a payment interface, a transaction interface, a user settings interface, and a history interface. The login interface is used for user login and identity verification. The payment interface is used for users to input or scan a code to select payee information and to display a preview of transaction details. The transaction interface is used to display the status of the transaction after it is initiated, show the payment path and transit node status, and can refresh transaction-related data in real time. The user settings interface is used for users to set personalized parameters such as payment methods, notification methods, and payment limits. The history interface is used to retrieve historical payment records and analyze historical payment paths.

[0045] The log recording terminal includes a recording and storage interface and a processing interface; the recording and storage interface is used to record and store user operation logs, transaction behavior logs, and payment operation logs; the processing interface is used to perform structured format processing and classification archiving of the logs from the storage interface.

[0046] The notification terminal is used to send payment status reminders, abnormal warnings, confirmation information, and security notifications to users;

[0047] The large-scale fund payment model is connected to the user interface for intelligent use by users, improving the efficiency and security of fund payment management.

[0048] By using deep learning to study fund application data, a large-scale fund payment model is constructed.

[0049] The large-scale payment model includes a data input module and a model application module. The data input module is used to input payment application data and continuously update the large-scale payment model. The model application module is used to analyze and apply the payment application data, including a path optimization model, a risk model, and a user profile model. The path optimization model uses a neural network model, combined with payment channels and transaction processes, to establish payment nodes and payment edges, connecting payment nodes through payment edges to construct a payment channel graph. Deep reinforcement learning is used to process the payment channel graph, optimizing it with the maximum success rate and minimum cost as training objectives. The risk model uses Transformer technology to analyze user transaction behavior, identify risky user behaviors, and store these as risk sample behaviors within the risk model. The user profile model acquires user feature data from the payment application data and payment channel feature data from the path optimization model. DeepFM is used to fuse and analyze the user feature data and payment channel feature data, and combined with the risk model, risky behavior data is eliminated, outputting user payment preferences and recommended payment paths.

[0050] It should be further explained that, in the specific implementation process, a risk assessment is conducted on the payment process by combining the fund recommendation path and the overall fund payment model. The risk assessment results are then used to manage the payment process, and the process of interacting with users through the user interface is as follows:

[0051] By comparing with the fund application data within the large fund payment model, path channel data and transaction behavior data for the fund recommendation path are obtained. The path channel data includes the number of payment failures and the total number of payments at each payment node. The transaction behavior data includes the number of abnormal transactions and the total number of transactions at each payment edge. The path factor of the payment node is obtained by calculating the ratio of the number of payment failures to the total number of payments at each payment node. The transaction factor of the payment edge is obtained by calculating the ratio of the number of abnormal transactions to the total number of transactions at each payment edge. Payment path weights and payment transaction weights are set. The path evaluation value is obtained by multiplying the payment path weights, payment transaction weights, path factors, and transaction factors.

[0052] Set risk threshold one, risk threshold two, risk threshold three, and risk threshold four;

[0053] When the path assessment value is less than or equal to the risk threshold one, the risk of the recommended funding path corresponding to the path assessment value is low risk. The recommended funding path can then be sent to the user interface for recommendation to the user, and a risk assessment result one can be obtained.

[0054] When the path assessment value is greater than risk threshold one and less than risk threshold two, the risk of the recommended funding path corresponding to the path assessment value is medium risk. The risk area on the recommended funding path can be marked and sent to the user interface to remind the user and suggest that the user enable secondary verification to regulate the user's payment behavior and obtain risk assessment result two.

[0055] When the path assessment value is greater than or equal to risk threshold two and less than risk threshold three, the risk of the recommended funding path corresponding to the path assessment value is high risk. The user will be notified to review and verify their identity to obtain risk assessment result three.

[0056] When the path assessment value is greater than or equal to risk threshold three and less than or equal to risk threshold four, the risk of the recommended funding path corresponding to the path assessment value is extremely high. In this case, the user account will be frozen and all transactions of the user account will be intercepted, resulting in risk assessment result four.

[0057] When risk assessment result one is generated, the recommended funding path corresponding to risk assessment result one will be sent to the user interface to guide the user to make payment transactions according to the recommended funding path;

[0058] When Risk Assessment Result 2, Risk Assessment Result 3, or Risk Assessment Result 4 is generated, the risk data of the recommended funding path corresponding to Risk Assessment Result 2, Risk Assessment Result 3, or Risk Assessment Result 4 will be sent to the user interface to remind the user not to make payment transactions based on the risk data.

[0059] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A fund routing payment system based on a large-scale model application, including an order center, characterized in that: The order center communication connection includes a data acquisition and processing module, a model building module, and a risk management module. The data acquisition and processing module is used to acquire fund payment data from multiple internal and external data sources, and to process the fund payment data to obtain fund application data. The model building module is used to set up the user interface and the large-scale fund payment model. Through the large-scale fund payment model, fund application data is analyzed to obtain fund recommendation paths. The risk management module is used to assess the risks of the payment process in conjunction with the recommended funding path, obtain the risk assessment results, manage the payment process based on the risk assessment results, and interact with users through the user interface.

2. The fund routing payment system based on large-scale model application according to claim 1, characterized in that, The process of obtaining payment data from multiple internal and external data sources includes: The multiple internal and external data sources include third-party payment platforms, merchant transaction systems, bank clearing interfaces, blockchain payment channels, user terminal behavior logs, and device fingerprint information; The payment data includes transaction data, payment channel data, and user device data; the transaction data includes transaction request data, transaction status data, historical transaction path data, and transaction purpose; the payment channel data includes payment method and congestion status; the user device data includes user identity information, device environment information, user preference data, user behavior data, and operation logs.

3. The fund routing payment system based on large-scale model application according to claim 2, characterized in that, The process of processing fund payment data to obtain fund application data includes: The fund payment data is refined, including data deduplication, data completion, data correction, and data unification; fund payment data belonging to the same payment process are grouped into one category and referred to as fund payment flow data; features are extracted from the fund payment flow data to obtain fund application data; the fund application data includes transaction feature data, payment feature data, and user feature data.

4. The fund routing payment system based on large-scale model application according to claim 3, characterized in that, The process of setting up the user interface is as follows: Set up a user interaction interface, which includes a user terminal, a log recording terminal, and a notification terminal; The user interface includes a login interface, a payment interface, a transaction interface, a user settings interface, and a history interface. The login interface is used for user login and identity verification. The payment interface is used for users to input or scan a code to select payee information and to display a preview of transaction details. The transaction interface is used to display the status of the transaction after it is initiated, show the payment path and transit node status, and can refresh transaction-related data in real time. The user settings interface is used for users to set personalized parameters such as payment method, notification method, and payment limit. The history interface is used to retrieve historical payment records and analyze historical payment paths. The log recording terminal includes a recording and storage interface and a processing interface; the recording and storage interface is used to record and store user operation logs, transaction behavior logs, and payment operation logs; the processing interface is used to perform structured format processing and categorized archiving of the logs from the storage interface. The notification terminal is used to send payment status reminders, abnormal warnings, confirmation information, and security notifications to users.

5. The fund routing payment system based on large-scale model application according to claim 4, characterized in that, The process of setting up a large-scale payment model includes: The large-scale fund payment model is connected to the user interface and is constructed by deep learning of fund application data. The large-scale fund payment model includes a data input module and a model application module; the data input module is used to input fund application data and continuously update the large-scale fund payment model; the model application module is used to analyze and apply the fund application data, including a path optimization model, a risk model, and a user profile model.

6. The fund routing payment system based on large-scale model application according to claim 5, characterized in that, The process of analyzing fund application data through a large-scale fund payment model to obtain recommended fund paths includes: The path optimization model uses a neural network model, combined with payment channels and transaction processes, to establish payment nodes and payment edges. Payment nodes are connected via payment edges to construct a payment channel graph. Deep reinforcement learning is used to process the payment channel graph, optimizing it with the maximum success rate and minimum cost as training objectives. The risk model uses Transformer technology to analyze user transaction behavior, identify risky user behaviors, and store these as risk sample behaviors within the risk model. The user profiling model acquires user feature data from the fund application data and payment channel feature data from the path optimization model. DeepFM is used to fuse and analyze the user feature data and payment channel feature data, and combined with the risk model, data exhibiting risky behaviors are eliminated, outputting user payment preferences and recommended fund paths.

7. The fund routing payment system based on large-scale model application according to claim 6, characterized in that, The process of assessing the risk of the payment process by combining the recommended funding path and the overall funding payment model, and obtaining the risk assessment results, includes: By comparing with the fund application data within the large fund payment model, path channel data and transaction behavior data of the fund recommendation path are obtained. The path channel data includes the number of payment failures and the total number of payments at each payment node. The transaction behavior data includes the number of abnormal transactions and the total number of transactions at each payment edge. Based on the path channel data and transaction behavior data, path factors and transaction factors are obtained. Payment path weights and payment transaction weights are set. Based on the payment path weights, payment transaction weights, path factors, and transaction factors, a path evaluation value is obtained. Set risk thresholds one, two, three, and four. When the path evaluation value is less than or equal to risk threshold one, obtain risk evaluation result one. When the path evaluation value is greater than risk threshold one, obtain risk evaluation result two. When the path evaluation value is greater than or equal to risk threshold two, obtain risk evaluation result three. When the path evaluation value is greater than or equal to risk threshold three, obtain risk evaluation result four.

8. The fund routing payment system based on large-scale model application according to claim 7, characterized in that, The process of managing the payment process based on risk assessment results and interacting with users through a user interface includes: When risk assessment result one is generated, the recommended funding path corresponding to risk assessment result one will be sent to the user interface to guide the user to make payment transactions according to the recommended funding path; When Risk Assessment Result 2, Risk Assessment Result 3, or Risk Assessment Result 4 is generated, the risk data of the recommended funding path corresponding to Risk Assessment Result 2, Risk Assessment Result 3, or Risk Assessment Result 4 will be sent to the user interface to remind the user not to make payment transactions based on the risk data.

Images