Method and system for automatically switching approval network based on payment card type analysis

The automatic approval network switching system addresses payment rejections by classifying card types and switching to a VAN network based on payment history, improving transaction efficiency and success rates.

KR103046675B1Active Publication Date: 2026-11-11ARIWORKS CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
KR1020260114534
Authority / Receiving Office
KR · KR
Patent Type
Patents
Current Assignee / Owner
Filing Date
2026-06-23
Publication Date
2026-11-11
Estimated Expiration
2046-06-23

AI Technical Summary

Technical Problem

Conventional payment processing technologies often reject payments or require manual intervention when a payment card is unsuitable for the configured approval network, fail to flexibly respond to new or special payment cards, and lack the ability to automatically switch to an appropriate network based on card type and payment history.

Method used

An automatic approval network switching system that analyzes payment card identification information to classify the card type, determines if payment is possible through a Payment Gateway (PG) network, and automatically switches to a Value Added Network (VAN) if necessary, using a classification model trained on payment history to improve switching accuracy and efficiency.

Benefits of technology

The system resolves payment rejections by automatically switching to a VAN network without manual intervention, enhances payment success rates, and continuously learns to handle new or special payment cards, ensuring seamless transaction processing.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 112026076306970-PAT00005_ABST
    Figure 112026076306970-PAT00005_ABST
Patent Text Reader

Abstract

A system for automatically switching authorization networks based on payment card type analysis may include a memory and a processor for storing instructions. The system may acquire card identification information from a payment card that is contacted, inserted, or tagged at a payment terminal, classify the type of payment card by analyzing the acquired card identification information, and determine whether payment is possible through a Payment Gateway (PG) authorization network based on the classified payment card type. If the system determines that payment through the Payment Gateway (PG) authorization network is not possible, it may automatically switch the authorization network for payment to a Value Added Network (VAN) authorization network, request payment authorization for the payment card through the switched VAN authorization network, and accumulate payment history data including the type of payment card, whether the authorization network has been switched, and the payment authorization result as training data, and control the system to update the classification of the payment card type and the determination of the authorization network switch based on the accumulated training data.
Need to check novelty before this filing date? Find Prior Art

Description

Technology Field

[0001] The present invention relates to the fields of payment processing and artificial intelligence.

[0002] More specifically, the invention relates to an automatic approval network switching method and system that analyzes identification information of a payment card presented on a payment terminal to classify the type of the payment card, automatically selects and switches to an approval network capable of payment among a Payment Gateway (PG) approval network and a Value Added Network (VAN) approval network based on the classified type, and learns payment history to update the switching rule. Background Technology

[0004] With the recent expansion of the electronic payment market, there is a growing trend of attempts to process various types of payment cards through payment terminals. As channels for processing payment approvals, there may be Payment Gateway (PG) approval networks via payment processing agencies and Value Added Network (VAN) approval networks via value-added network operators, and depending on the type of payment card and the issuer, there may be cases where payment is only possible through one of these approval networks.

[0005] However, existing payment processing technologies often relied on payment terminals requesting approval only through a pre-configured single approval network, or on operators manually specifying the approval path based on the type of payment card. Under these methods, there was the inconvenience of payments being rejected if a card unsuitable for the Fiji approval network was presented, or requiring the use of a different terminal or separate operations to complete the transaction.

[0006] Meanwhile, some conventional technologies for identifying payment cards and processing transactions focused on the terminal recognizing specific files or identification information set on the card and executing the transaction according to predetermined processing procedures. However, these conventional technologies were limited to distinguishing cards based on predefined fixed rules and failed to present configurations that analyze the type of payment card to automatically determine and switch to an authorized network for payment, or to learn payment history to autonomously update switching rules.

[0007] Furthermore, while there may be special payment cards in the market whose support for specific approval networks fluctuates frequently due to changes in issuer policies or the launch of new cards, it has been practically difficult to register and manage the approval paths for such cards individually in advance. The problem to be solved

[0009] Conventional payment processing technologies had problems such as rejecting payments or requiring manual intervention by operators when a payment card that cannot be used on the Fiji approval network was presented, and failing to flexibly respond to new or special payment cards.

[0010] The embodiments of the present invention aim to solve various problems, including the problems mentioned above, by analyzing identification information of a payment card presented to a payment terminal to classify the type of the payment card, and determining whether payment is possible through a Fiji approval network based on the classified type, and automatically switching to a VAN approval network if payment is not possible, thereby providing a method and system for automatically switching approval networks to complete payment without operator intervention.

[0011] In addition, the present invention aims to provide an automatic approval network switching method and system that improves switching accuracy by accumulating payment history data as training data and predicting the probability of payment possible through the Fiji approval network using a classification model trained thereon. This allows for the gradual learning of types of payment cards that cannot be used for payment in the Fiji approval network.

[0012] In addition, the present invention aims to provide an automatic approval network switching method and system that increases the success rate and efficiency of payments by selecting a suitable van approval network by considering status information including the response delay time and approval failure rate of each of a plurality of van approval networks, and switching to the next-ranked van approval network when payment to the selected van approval network fails.

[0013] In addition, the present invention aims to provide an automatic approval network switching method and system that automatically expands and keeps up-to-date switching rules for new or special payment cards by constructing a table mapping switching rules of the approval network according to the identification information of the payment card, registering the search results for unregistered cards as new items, and re-verifying the items according to changes in issuer policies.

[0014] However, these tasks are exemplary and do not limit the scope of the invention. means of solving the problem

[0016] An automatic authorization network switching system based on payment card type analysis may include a memory for storing instructions and a processor. When instructions are executed by the processor, the system may obtain card identification information from a payment card that is contacted, inserted, or tagged at a payment terminal, analyze the obtained card identification information to classify the type of payment card, and determine whether payment is possible through a Payment Gateway (PG) authorization network based on the classified payment card type. If the system determines that payment through the Payment Gateway (PG) authorization network is not possible, it may automatically switch the authorization network for payment to a Value Added Network (VAN) authorization network, request payment authorization for the payment card through the switched VAN authorization network, and accumulate payment history data including the type of payment card, whether the authorization network has been switched, and the payment authorization result as training data, and control the system to update the payment card type classification and authorization network switching determination based on the accumulated training data. Effects of the invention

[0018] According to an embodiment of the present invention, an approval network automatic switching method and system can be implemented that automatically analyzes the type of payment card and automatically selects and switches to an approval network capable of payment among a Fiji approval network and a VAN approval network.

[0019] In addition, the present invention has the effect of resolving the inconvenience caused by payment rejection and improving the convenience and speed of payment processing by automatically switching to a VAN approval network and completing the payment without manual intervention by an operator, even when a payment card that cannot be paid to a Fiji approval network is presented.

[0020] In addition, the present invention has the effect of accurately predicting whether to convert even for payment cards not previously defined and continuously improving conversion accuracy by progressively learning the types of payment cards that cannot be used for payment using a classification model trained on payment history data.

[0021] In addition, the present invention has the effect of increasing the payment success rate and flexibly responding to failures or delays in specific approval networks by considering the status information of a plurality of approval networks, selecting a suitable approval network, and switching to the next-ranked network in the event of a payment failure. Brief explanation of the drawing

[0023] FIG. 1 is a block diagram showing the configuration of an approval network automatic switching system based on payment card type analysis according to one embodiment. FIG. 2 is a flowchart illustrating an approval network automatic switching method according to one embodiment. FIG. 3 is a flowchart illustrating a method for learning and converting payment availability according to one embodiment. FIG. 4 is a flowchart illustrating a method for updating transition rules based on a transition mapping table according to one embodiment. FIG. 5 conceptually illustrates the overall operation flow of an approval network automatic switching system based on payment card type analysis according to one embodiment. Specific details for implementing the invention

[0024] The terms used in the embodiments are for illustrative purposes only and should not be interpreted in a restrictive sense. Singular expressions are understood to include plural forms unless the context clearly indicates otherwise. In this specification, the terms "comprising" or "having" mean that the features, figures, steps, actions, components, parts, or combinations thereof mentioned in the specification are present, and do not preclude the possibility that other features, figures, steps, actions, components, parts, etc. may exist.

[0025] Unless specifically defined otherwise, all terms used herein, including technical or scientific terms, shall be interpreted as having the same meaning as generally accepted in the relevant technical field. Terms defined in commonly used dictionaries shall be interpreted as having meanings consistent with the context of the relevant technology and shall not be interpreted in an overly formal or ideal sense unless specifically defined in this application.

[0026] Additionally, when describing with reference to the attached drawings, identical components are given the same reference numerals regardless of the drawing symbols, and redundant descriptions are omitted. When describing embodiments, if it is determined that a detailed description of related prior art could unnecessarily complicate the main contents of the embodiment, such description may be omitted.

[0028] FIG. 1 is a block diagram showing the configuration of an approval network automatic switching system based on payment card type analysis according to one embodiment.

[0029] Referring to FIG. 1, the system (100) may include a communication circuit (110), a processor (120), a memory (130), and a learning engine (140). The configuration shown in FIG. 1 is merely an example and is not limited thereto, and the system (100) may omit some of the illustrated configurations or include additional configurations that are not illustrated. Each configuration of the system (100) may be implemented integrated within a single device or distributed across multiple devices.

[0030] The communication circuit (110) can enable the system (100) to transmit and receive data with an external device. The communication circuit (110) can receive card identification information and payment information of a payment card from a payment terminal. The communication circuit (110) can transmit and receive payment approval requests and responses with a Payment Gateway (PG) approval network and a Payment Network (VAN) approval network. The communication circuit (110) can support at least one of a wired communication method or a wireless communication method. The communication circuit (110) can be connected to each of a plurality of approval networks simultaneously or sequentially. The communication circuit (110) can physically perform a switching operation to terminate a payment session to one approval network and create a payment session to another approval network. The types of communication methods are merely examples and are not limited thereto, and may vary depending on the settings.

[0031] The processor (120) can control the overall operation of the system (100). The processor (120) can execute instructions stored in memory (130) to perform analysis of card identification information, classification of payment card types, determination of whether payment is possible through the Fiji approval network, and automatic switching to the VAN approval network. The processor (120) can analyze the card identification information received by the communication circuit (110) to extract BIN information, issuer identifier, and card brand identifier. The processor (120) can classify the type of payment card by comparing the extracted information with a BIN range table. The processor (120) can determine whether to switch to the approval network and the target of the switch based on the classification results and the inference results of the learning engine (140). The processor (120) can calculate priority scores for multiple VAN approval networks and select the most suitable VAN approval network. The processor (120) can be implemented as one or more central processing units or microprocessors.

[0032] Memory (130) can store data and instructions necessary for the operation of the system (100). Memory (130) can store instructions executed by the processor (120). Memory (130) can store a blank range table, a list of issuers supported by the Fiji approval network, network correspondence information matching card types with the van approval network by issuer, and a conversion mapping table. Memory (130) can store payment history data and training data used for training a classification model. Memory (130) can store weight parameters of the trained classification model. Memory (130) can function as a data storage for the system (100) in place of a separate database. Memory (130) may include at least one of volatile memory or non-volatile memory.

[0033] The learning engine (140) can perform training and inference of a classification model that predicts payment eligibility. The learning engine (140) can be implemented as a separate computing device from the processor (120), or as a software module executed by the processor (120). The learning engine (140) can train the classification model using training data stored in memory (130).

[0034] The learning engine (140) may configure the learning data into input items and correct values. The input items may include empty information, an issuer identifier, a card brand identifier, and a payment request amount. The correct value may indicate whether the payment card is eligible for payment through the Fiji approval network. The learning engine (140) may perform preprocessing for empty information and issuer identifiers, which do not have meaning in order or size, by setting separate input slots for each type, displaying 1 only in the corresponding slots, and displaying 0 in the remaining slots. The learning engine (140) may perform preprocessing for the payment request amount by converting the value between a preset minimum value and maximum value into a value between 0 and 1 using a ratio.

[0035] The learning engine (140) can input preprocessed training data into a classification model to output a payment probability through the Fiji approval network as a value between 0 and 1. The classification model can be implemented as any one of a logistic regression model, a decision tree model, or a multilayer neural network model. The learning engine (140) can calculate the difference between the payment probability output by the classification model and the correct value as an error. The learning engine (140) can repeatedly adjust the internal weights of the classification model in a direction that reduces the calculated error. The learning engine (140) can repeat the weight adjustment until the error converges to a preset level or reaches a preset number of iterations. The types of classification models are merely examples and are not limited to this, and may vary depending on the settings.

[0036] The learning engine (140) can perform inference using a learned classification model. The learning engine (140) can input card identification information of a newly presented payment card into the classification model to calculate the probability of payment possible through the Fiji approval network. The learning engine (140) can transmit the calculated payment probability to the processor (120). The processor (120) can determine whether to convert by comparing the received payment probability with a preset threshold probability.

[0037] The learning engine (140) can update the classification model using payment results. The learning engine (140) can add training data to memory (130) in which the payment approval results of the ban approval network based on automatic conversion are the correct answers. The learning engine (140) can retrain the classification model whenever the added training data exceeds a preset number. Through this retraining, the learning engine (140) can gradually learn the types of payment cards that cannot be used for payment in the Fiji approval network. The learning engine (140) can store the updated weight parameters in memory (130) and reflect them in subsequent inference.

[0039] FIG. 2 is a flowchart illustrating an approval network automatic switching method according to one embodiment.

[0040] The operations described through FIG. 2 can be implemented based on instructions that can be stored in memory (e.g., memory (130) of FIG. 1). The illustrated method can be executed by the system (100), the order of each operation of FIG. 2 can be changed, some operations can be omitted, and some operations can be performed simultaneously.

[0041] In operation 210, the system (100) can obtain card identification information from the payment card and classify the type of the payment card.

[0042] According to one embodiment, the system (100) may obtain card identification information from a payment card that is contacted, inserted, or tagged in a payment terminal. The card identification information may include bin information corresponding to the leading digit of the card number, an issuer identifier, and a card brand identifier. The system (100) may read the card identification information through at least one of contact communication, contactless communication, or magnetic stripe reading. The system (100) may compare the obtained bin information with a pre-stored bin range table. The bin range table may refer to a table that records card types corresponding to each interval of the bin information. Based on the comparison result, the system (100) may classify the payment card into any one of a credit card, debit card, prepaid card, corporate card, and special purpose card. The types and number of card types are merely examples and are not limited thereto, and may vary depending on the settings.

[0043] In operation 220, the system (100) can determine whether payment is possible through the Fiji approval network based on the classified type.

[0044] According to one embodiment, the system (100) may compare classified type and issuer identifiers with a list of issuers supported by the Fiji approval network. If the payment card is included in the list of issuers, the system (100) may send a payment approval request to the Fiji approval network. The system (100) may check the response code received from the Fiji approval network. If the response code is a predefined normal approval code, the system (100) may determine that payment through the Fiji approval network is possible. If the response code is a rejection response code corresponding to issuer non-support, card type non-support, or transaction rejection, the system (100) may determine that payment through the Fiji approval network is not possible. Additionally, the system (100) may determine that payment through the Fiji approval network is not possible if the payment card is not included in the list of issuers or if a response code is not received within a pre-set waiting time.

[0045] In operation 230, if payment is not possible, the system (100) can automatically switch to a van approval network and request payment approval.

[0046] According to one embodiment, if the system (100) determines that payment through the Fiji approval network is not possible, it may look up and select a VAN approval network corresponding to the type of payment card and issuer identifier from pre-stored network correspondence information. The system (100) may terminate the payment session to the Fiji approval network that was in progress and create a payment session with the selected VAN approval network. The system (100) may convert payment information obtained from the payment card into a message format of the VAN approval network. The system (100) may transmit the converted message to the VAN approval network to request payment approval. The system (100) may automatically switch the approval network and complete the payment without manual intervention by an operator.

[0047] According to one embodiment, an automatic approval network switching system (100) based on payment card type analysis may include a memory and a processor for storing instructions. When instructions are executed by the processor, the system (100) may obtain card identification information from a payment card that is contacted, inserted, or tagged at a payment terminal, analyze the obtained card identification information to classify the type of the payment card, and determine whether payment is possible through a Payment Gateway (PG) approval network based on the classified type of payment card. If the system (100) determines that payment through the Payment Gateway (PG) approval network is not possible, it may automatically switch the approval network for payment to a Payment Network (VAN) approval network, request payment approval for the payment card through the switched VAN approval network, and accumulate payment history data including the type of payment card, whether the approval network has been switched, and the payment approval result as training data, and control the system to update the payment card type classification and approval network switching determination based on the accumulated training data.

[0048] The system (100) can automatically switch the approval network by analyzing the identification information of a payment card presented at a payment terminal. Here, the approval network may refer to a communication path through which a payment approval request is transmitted and processed. The approval network may include a Payment Gateway (PG) approval network passing through a payment agency and a Value Added Network (VAN) approval network passing through a Value Added Network operator. The PG approval network may be a path where a payment agency acts as an intermediary for linking with multiple card companies in an online payment environment. The VAN approval network may be a path where a Value Added Network operator acts as an intermediary for approval between a card company and a merchant in an offline payment environment. Even with the same payment card, there may be cases where payment is possible only through one of the approval networks depending on the type of card or the issuer.

[0049] The system (100) can first classify the types of payment cards to address these differences. The types of payment cards may refer to classifications based on the issuance form or purpose of the card. The system (100) can determine which type a payment card belongs to by analyzing card identification information. Based on the classified types, the system (100) can determine whether payment is possible through the Fiji approval network. If the system (100) determines that payment is not possible through the Fiji approval network, it can switch to the VAN approval network. This switching can be performed automatically by the system (100) without operator intervention. The system (100) can reduce situations where payment is interrupted due to payment rejection.

[0050] The system (100) can accumulate payment history data as training data. Payment history data may refer to data including the card type, approval network conversion status, and payment approval result of each processed payment transaction. The system (100) can accumulate and analyze this data to update type classification criteria and conversion judgment criteria. The system (100) can increase the accuracy of conversion judgment as new payment cards are presented.

[0051] According to one embodiment, the system (100) obtains bin information corresponding to the leading digit of the card number, an issuer identifier, and a card brand identifier from a payment card as card identification information, and compares the obtained bin information with a pre-stored bin range table to classify the payment card into one of the types of a credit card, a debit card, a prepaid card, a corporate card, and a special purpose card. The system (100) first checks whether a payment card is included in the list by comparing a classified type and issuer identifier with a list of issuers supported by the Fiji approval network to determine whether payment is possible through the Fiji approval network, and if the payment card is included in the list, it sends a payment approval request to the Fiji approval network and checks the received response code. If the response code is a predefined normal approval code, it determines that payment is possible through the Fiji approval network, and if the response code is a rejection response code corresponding to issuer not supported, card type not supported, or transaction rejection, it determines that payment is not possible through the Fiji approval network. It may also determine that payment is not possible through the Fiji approval network if the payment card is not included in the list or if a response code is not received from the Fiji approval network within a pre-set waiting time.

[0052] According to one embodiment, the system (100) can control the automatic switching of the approval network by, when it is determined that payment through the Fiji approval network is not possible, querying and selecting a VAN approval network corresponding to the type of payment card and issuer identifier from pre-stored network correspondence information, terminating the payment session to the Fiji approval network that was in progress, creating a payment session with the selected VAN approval network, and converting the payment information obtained from the payment card into the message format of the VAN approval network to request payment approval.

[0053] According to one embodiment, the system (100) may obtain bin information as card identification information. Bin information may refer to an identification number corresponding to the leading digit of a card number. Bin information may be used to identify the institution that issued the card and the type of card. Along with the bin information, the system (100) may obtain an issuer identifier and a card brand identifier. An issuer identifier may refer to information that distinguishes the financial institution that issued the card. A card brand identifier may refer to information that distinguishes the payment network brand of the card.

[0054] According to one embodiment, the system (100) can compare the acquired empty information with an empty range table. An empty range table may refer to a table in which card types corresponding to each section of empty information are pre-recorded. The system (100) can classify the type of payment card by identifying the section to which the empty information belongs. The types and number of card types are merely examples and are not limited thereto, and may vary depending on the settings.

[0055] According to one embodiment, the system (100) can determine whether payment is possible in stages. The system (100) can first compare classified type and issuer identifiers with a list of issuers. The list of issuers may refer to a list of issuers that the Fiji approval network has organized in advance to support payments. If the payment card is included in the list of issuers, the system (100) may determine that payment is possible in the first instance and send a payment approval request to the Fiji approval network. The system (100) may check the response code thereon. The response code may refer to a code indicating the result of processing the approval request. The system (100) may determine that payment is possible if the response code is a normal approval code. The system (100) may determine that payment is not possible if the response code is a rejection response code corresponding to issuer non-support, card type non-support, or transaction rejection.

[0056] According to one embodiment, the system (100) may determine that payment is not possible even if the payment card is not included in the list of issuers. The system (100) may also determine that payment is not possible even if a response code is not received within a waiting time. Through these determinations, the system (100) may include list comparison, response verification, and non-response processing as reasons for payment impossibility. The system (100) may reduce the number of omitted determination reasons.

[0057] According to one embodiment, the system (100) may switch to a VAN approval network if it is determined that payment is not possible. The system (100) may query a VAN approval network corresponding to the payment card type and issuer identifier from network correspondence information. Network correspondence information may refer to information in which a suitable VAN approval network is pre-matched for each card type and issuer. The system (100) may terminate a payment session to a Fiji approval network that was in progress. A payment session may refer to a unit of communication maintained in connection with an approval network to process a single payment transaction. The system (100) may create a new payment session with a selected VAN approval network. The system (100) may convert payment information into a message format of the VAN approval network. A message format may refer to the structure and specifications of data required by the corresponding approval network. The system (100) may request payment approval by transmitting the converted message. The system (100) can process the same payment transaction to another approval network without replacing a separate terminal by going through session termination, session creation, and professional conversion.

[0059] FIG. 3 is a flowchart illustrating a method for learning and converting payment availability according to one embodiment.

[0060] The operations described through FIG. 3 can be implemented based on instructions that can be stored in memory (e.g., memory (130) of FIG. 1). The illustrated method can be executed by the system (100), the order of each operation of FIG. 3 can be changed, some operations can be omitted, and some operations can be performed simultaneously.

[0061] In operation 310, the system (100) can configure training data with card identification information and payment availability.

[0062] According to one embodiment, the system (100) may configure training data for classifying the types of payment cards. The training data may refer to data in which card identification information, such as empty information, an issuer identifier, a card brand identifier, and a payment request amount, are input items, and whether the corresponding payment card was able to make a payment through the Fiji approval network is set as the correct answer value. The system (100) may extract the card identification information and the actual payment availability of each payment card from payment history processed in the past and accumulate them as training data. The system (100) may expand the amount of training data as payments accumulate.

[0063] In operation 320, the system (100) can perform preprocessing to transform and normalize input items.

[0064] According to one embodiment, the system (100) can perform preprocessing on input items included in the training data. For items where the order does not have meaning, such as empty information and issuer identifiers, the system (100) can convert them by placing a separate input position for each type of item, displaying 1 only in the corresponding position, and displaying 0 in the remaining positions. For items that have a size, such as payment request amounts, the system (100) can convert them by a ratio so that the value between a preset minimum value and a maximum value becomes a value between 0 and 1. Through such preprocessing, the system (100) can convert different types of input items into a numerical form that a classification model can process.

[0065] In operation 330, the system (100) can train the classification model by repeatedly adjusting its weights so that the error becomes smaller.

[0066] According to one embodiment, the system (100) may input preprocessed training data into a classification model. The classification model may refer to a model that outputs a payment possibility probability through a Fiji approval network from input card identification information as a value between 0 and 1. The classification model may be implemented as any one of a logistic regression model, a decision tree model, or a multilayer neural network model. The system (100) may calculate the difference between the payment possibility probability output by the classification model and the correct answer value indicating whether actual payment is possible as an error. The system (100) may repeatedly adjust the internal weights of the classification model in a direction that reduces the calculated error. The system (100) may repeat the weight adjustment until the error converges to a preset level or reaches a preset number of iterations.

[0067] In operation 340, the system (100) can predict the probability of payment possible, convert it, and retrain the model as a result.

[0068] According to one embodiment, the system (100) may input card identification information of a newly contacted, inserted, or tagged payment card into a learned classification model. The system (100) may compare the probability of payment possible through the Fiji approval network calculated by the classification model with a preset threshold probability. If the probability of payment possible is less than the threshold probability, the system (100) may automatically switch the approval network for payment to a VAN approval network. The system (100) may add training data to the existing training data, with the payment approval result of the VAN approval network resulting from the automatic switch being the correct answer. The system (100) may retrain the classification model using the added training data. Through this retraining, the system (100) may gradually learn the types of payment cards that cannot be used for payment in the Fiji approval network.

[0069] The system (100) can train a classification model to classify the types of payment cards. A classification model may refer to a model trained to output the probability of a specific result from input information. The system (100) may configure training data to train the classification model. The training data may consist of input items and correct answer values. Input items may include empty information, an issuer identifier, a card brand identifier, and a payment request amount. The correct answer value may indicate whether the corresponding payment card was eligible for payment through the Fiji approval network. The system (100) may extract input items and correct answer values ​​from payment history processed in the past and accumulate them as training data.

[0070] According to one embodiment, the system (100) may perform preprocessing on the training data. Preprocessing may refer to the process of converting input items into a numerical form that can be processed by a classification model. Empty information or issuer identifiers may be items that do not have a meaning of order or size. The system (100) may convert these items by placing a separate input position for each type, displaying 1 only in the corresponding position, and displaying 0 in the remaining positions. Through this conversion, the system (100) can prevent the formation of unnecessary size relationships between items. The payment request amount may be an item that has a meaning of size. The system (100) may convert the payment request amount into a ratio such that the value between the minimum and maximum values ​​becomes a value between 0 and 1. Through this conversion, the system (100) can prevent the amount item from having an excessively large influence compared to other items.

[0071] According to one embodiment, the system (100) can train a classification model using preprocessed training data. The classification model can output a payment possibility probability between 0 and 1 through a Fiji approval network from input card identification information. The classification model can be implemented as any one of a logistic regression model, a decision tree model, or a multilayer neural network model. The system (100) can calculate the difference between the payment possibility probability output by the classification model and the correct answer value as an error. An error may refer to a value indicating how much the model's prediction deviates from the actual correct answer. The system (100) can iteratively adjust the internal weights of the classification model in a direction that reduces the calculated error. A weight may refer to a value that determines the degree of influence an input item has on the output. The system (100) can repeat the adjustment until the error converges to a preset level or reaches a preset number of iterations. Through such iterations, the system (100) can obtain a model that predicts payment possibility more accurately.

[0072] According to one embodiment, the system (100) can utilize a learned classification model for actual payment processing. The system (100) can input card identification information of a newly presented payment card into the classification model. The system (100) can compare the calculated payment probability with a threshold probability. The threshold probability may refer to a probability value that serves as a criterion for determining whether to convert. If the payment probability is less than the threshold probability, the system (100) can convert to a VAN approval network. The system (100) can accumulate training data in which the payment approval result of the VAN approval network following the conversion is used as the correct answer value. The system (100) can retrain the classification model using the accumulated training data. Through this retraining, the system (100) can gradually respond to new payment card types that are not previously defined. The system (100) can improve the accuracy of conversion judgment as payments accumulate.

[0073] According to one embodiment, the system (100) may construct training data for classifying the type of payment card, with card identification information including empty information, issuer identifier, card brand identifier, and payment request amount as input items, and whether the payment card was able to be paid to the Fiji approval network as the correct answer value, and for the constructed training data, items for which the order does not have meaning, such as empty information and issuer identifier, are converted by placing a separate input digit for each type of item, displaying 1 only in the corresponding digit and 0 in the remaining digits, and the payment request amount may be converted into a ratio such that the value between a preset minimum value and a maximum value becomes a value between 0 and 1.

[0074] According to one embodiment, the system (100) trains a classification model that outputs a payment possibility probability through a Fiji approval network as a value between 0 and 1 from input card identification information using preprocessed training data, and trains the classification model by calculating the difference between the payment possibility probability output by the classification model and the correct answer value indicating whether actual payment is possible as an error, and repeatedly adjusting the internal weights of the classification model in a direction where the calculated error becomes smaller, and controls the classification model to gradually learn the types of payment cards that cannot be paid in the Fiji approval network by inputting card identification information of a newly contacted, inserted, or tagged payment card into the trained classification model and, when the payment possibility probability through the Fiji approval network calculated is less than a preset threshold probability, automatically switching the approval network for payment to a VAN approval network, and adding training data in which the payment approval result of the VAN approval network resulting from the automatic switching is the correct answer value to the training data and retraining the classification model.

[0075] According to one embodiment, the system (100) collects and periodically updates network status information for each of a plurality of VAN approval networks, including a response delay time, which means the time from the time of the payment request to the time of receiving the approval response; an approval failure rate, which means the ratio of the number of rejections or no responses to the total number of approval requests during a preset unit period; whether the payment card type is supported; and a fee charged per payment approval. When it is determined that payment through the VAN approval network is not possible with the payment card, the system (100) can select VAN approval networks that support the payment card type among the plurality of VAN approval networks as candidates. For each of the selected candidates, the system (100) calculates a priority score by summing the values ​​obtained by multiplying the response delay time, the approval failure rate, and the fee by preset weights, such that the shorter the response delay time, the lower the approval failure rate, and the lower the fee, the higher the priority score is. The system can then select the VAN approval network with the highest priority score as the approval network for payment and automatically switch it.

[0076] According to one embodiment, the system (100) may, when an approval response is not received within a preset waiting time or a rejection response code is received for a payment approval request through a selected van approval network, switch to a van approval network with a lower priority score to request payment approval again, and sequentially repeat switching to a lower priority van approval network until payment is approved or until all candidates are exhausted, and may update network status information by accumulating payment approval results and response delay times for each candidate to correct the approval failure rate and response delay time reflected in the calculation of the priority score in real time, and control to exclude the corresponding van approval network from the candidates during a preset recovery period if the approval failure rate of a specific van approval network exceeds a preset blocking threshold.

[0077] According to one embodiment, the system (100) can select one suitable for payment from among a plurality of van approval networks. There may be a plurality of van approval networks in the market. Each van approval network may have different characteristics in terms of processing speed, stability, and fees. The system (100) can select a better van approval network for each payment by reflecting these differences.

[0078] According to one embodiment, the system (100) may collect network status information for each VAN approval network. Network status information may refer to information indicating the payment processing performance and costs of each VAN approval network. Network status information may include response delay time, approval failure rate, support status by payment card type, and a fee charged per payment approval. Response delay time may refer to the time from the time of the payment request to the time of receiving the approval response. The shorter the response delay time, the faster the payment can be processed. Approval failure rate may refer to the ratio of the number of rejections or no response to the total number of approval requests during a unit period. The lower the approval failure rate, the more stably the payment can be processed. The system (100) may periodically update this status information. Through updating, the system (100) may reflect the latest status of each VAN approval network.

[0079] According to one embodiment, the system (100) can select candidates when it is determined that payment is not possible. The system (100) can select only the van approval networks that support the type of payment card among a plurality of van approval networks as candidates. The system (100) can reduce unnecessary attempts by excluding van approval networks that do not support it in advance.

[0080] According to one embodiment, the system (100) can calculate a priority score for each selected candidate. The priority score may represent a numerical value indicating how suitable each candidate is for payment processing. The system (100) can calculate the priority score by summing the values ​​obtained by multiplying the response delay time, approval failure rate, and fee by a preset weight for each. The system (100) can calculate the score such that a higher score is assigned as the response delay time is shorter, the approval failure rate is lower, and the fee is lower. The weights can determine the degree of influence each item has on the score. The values ​​of the weights are merely examples and are not limited thereto, and may vary depending on the settings. The system (100) can select and switch to the van approval network with the highest priority score. The system (100) can make a balanced choice by considering speed, stability, and cost together.

[0081] According to one embodiment, the system (100) can prepare for the case where payment to a selected van approval network fails. The system (100) may switch to a next-ranked van approval network if an approval response is not received within a waiting time or if a rejection response code is received. The system (100) may sequentially repeat switching to a next-ranked van approval network until payment is approved or until all candidates are exhausted. Through this sequential switching, the system (100) can increase the likelihood of completing payment even if a failure occurs in one van approval network.

[0082] The system (100) can reflect payment results back into the status information. The system (100) can update network status information by accumulating payment approval results and response delay times for each candidate. The system (100) can reuse the updated status information for calculating subsequent priority scores. Through this cycle, the system (100) can correct the approval failure rate and response delay time in real time. Additionally, if the approval failure rate of a specific van approval network exceeds a preset blocking threshold, the system (100) can exclude that van approval network from the candidates during a preset recovery period. The blocking threshold may refer to a standard value for determining whether to temporarily exclude the network. The recovery period may refer to the waiting period before the excluded network is included back in the candidates. The system (100) can reduce repeated failures by excluding networks with poor status for a certain period.

[0084] FIG. 4 is a flowchart illustrating a method for updating transition rules based on a transition mapping table according to one embodiment.

[0085] The operations described through FIG. 4 can be implemented based on instructions that can be stored in memory (e.g., memory (130) of FIG. 1). The illustrated method can be executed by the system (100), the order of each operation of FIG. 4 can be changed, some operations can be omitted, and some operations can be performed simultaneously.

[0086] In operation 410, the system (100) can store a conversion mapping table for each card identification information.

[0087] According to one embodiment, the system (100) may store a conversion mapping table for each payment card's VAN information and issuer identifier. The conversion mapping table may refer to a table that records the possibility of payment through a Fiji approval network, a suitable VAN approval network identifier, and the update time of the corresponding item. The system (100) may store the conversion mapping table in a database format and query it at the time of payment processing. By referring to the conversion mapping table, the system (100) can quickly determine the conversion rules for each payment card.

[0088] In operation 420, the system (100) can determine whether to switch to a table reference and the target for the registration card.

[0089] According to one embodiment, the system (100) can check whether the card identification information of a payment card is registered in a conversion mapping table. If the card identification information is registered in the conversion mapping table, the system (100) can omit the payment attempt to the Fiji approval network. The system (100) can determine whether the approval network is converted and the target approval network by referring to the conversion mapping table. The system (100) can reduce the time required for payment processing by omitting unnecessary payment attempts for the registered card.

[0090] In operation 430, the system (100) can search for an approval network capable of making payments for an unregistered card.

[0091] According to one embodiment, the system (100) can determine whether the payment card is an unregistered card whose card identification information is not registered in the conversion mapping table. If the card is unregistered, the system (100) may first attempt payment through the Fiji approval network. When the system (100) receives a rejection response code from the Fiji approval network, it may attempt payment sequentially to multiple van approval networks. Through sequential attempts, the system (100) may search for an approval network capable of making payment for the unregistered card.

[0092] In operation 440, the system (100) can register the search result as a new item in the table.

[0093] According to one embodiment, the system (100) can register the search results of a payment-enabled approval network for an unregistered card in a conversion mapping table. The system (100) can register the unregistered card's blank information, issuer identifier, whether payment is possible through the Fiji approval network, and the VAN approval network identifier of the successful payment as new items. By registering new items, the system (100) can apply conversion rules without going through the search process when a payment card with the same blank information is presented thereafter.

[0094] In operation 450, the system (100) can reset and re-verify items based on the accumulated failures and validity period.

[0095] According to one embodiment, the system (100) can aggregate the cumulative number of payment failures to an existing registered VAN approval network for a payment card having the same bin information. If the cumulative number of payment failures exceeds a preset threshold, the system (100) can reset the target VAN approval network for the item corresponding to the bin information to another VAN approval network that has successfully made a payment. Additionally, the system (100) can check whether a preset validity period has elapsed from the time of update for each item in the conversion mapping table. The system (100) can mark items whose validity period has expired as targets for verification by retrying payment to the Fiji approval network. The system (100) can keep the conversion rules up to date even if the support status of the Fiji approval network changes due to a change in the issuer's policy.

[0096] According to one embodiment, the system (100) stores a transition mapping table that corresponds the possibility of payment through the Fiji approval network, a suitable VAN approval network identifier, and the time of update of the corresponding item to the BIN information and issuer identifier of the payment card, and if the card identification information of the payment card is registered in the transition mapping table, the system can refer to the transition mapping table to skip the attempt to make a payment to the Fiji approval network and immediately determine whether to transition the approval network and the VAN approval network to be transitioned. In the case of an unregistered card where the card identification information of the payment card is not registered in the transition mapping table, the system (100) first attempts to make a payment through the Fiji approval network for the unregistered card, and if a rejection response code is received as a result, attempts to make a payment sequentially to a plurality of VAN approval networks to search for an approval network capable of making a payment, and as a result of searching for an approval network capable of making a payment for the unregistered card, the BIN information of the unregistered card, the issuer identifier, the possibility of payment through the Fiji approval network, and the VAN approval network identifier of the successful payment can be registered as new items in the transition mapping table. The system (100) can control the conversion of the conversion target network for a new or special payment card to be automatically expanded and kept up to date even when the support status of the Fiji approval network changes due to changes in the issuer's policy, by resetting the conversion target network for the item corresponding to the same empty information to another network that has successfully made a payment when the accumulated number of payment failures to the existing registered VAN approval network for a payment card having the same empty information exceeds a preset threshold, and when the preset validity period has elapsed from the time of update for each item in the conversion mapping table, marking the item as a target for verification by retrying payment to the Fiji approval network.

[0097] According to one embodiment, the system (100) may manage conversion rules using a conversion mapping table. A conversion mapping table may refer to a table that records conversion rules for each identification information of a payment card. The system (100) may store the possibility of payment through the Fiji approval network, a suitable VAN approval network identifier, and the update time of the corresponding item by correlating them with VAN information and issuer identifier. The update time may refer to the time when the item was last registered or changed. The system (100) may determine how old an item is by recording the update time.

[0098] According to one embodiment, the system (100) may perform different processing depending on whether the payment card is registered in the table. If the payment card identification information is registered in the conversion mapping table, the system (100) may omit the attempt to make a payment to the Fiji approval network. The system (100) may refer to the table to immediately determine whether the approval network is converted and the target approval network for conversion. Through this omission, the system (100) may not repeat unnecessary Fiji approval requests for cards that have already been confirmed as unusable for payment. The system (100) may reduce the time required for payment processing.

[0099] According to one embodiment, the system (100) may process a payment card that is not registered as an unregistered card. An unregistered card may refer to a payment card in which card identification information is not recorded in a conversion mapping table. The system (100) may first attempt payment through a Fiji approval network for the unregistered card. When a rejection response code is received from the Fiji approval network, the system (100) may sequentially attempt payment through multiple van approval networks. Through these sequential attempts, the system (100) may search for an approval network capable of making payment for the unregistered card. The system (100) may find a suitable path through actual attempts, even for a card for which there is no prior information.

[0100] According to one embodiment, the system (100) can reflect the search results in a table. The system (100) can register the blank information of an unregistered card, the issuer identifier, whether payment is possible through the Fiji approval network, and the VAN approval network identifier of a successful payment as new items. By registering new items, the system (100) can skip the search and apply a conversion rule when a payment card with the same blank information is presented thereafter. The system (100) can expand the conversion rule itself whenever a new card is processed.

[0101] According to one embodiment, the system (100) may continuously update registered items. The system (100) may aggregate the cumulative number of payment failures to previously registered VAN approval networks for a payment card having the same empty information. If the cumulative number of payment failures exceeds a preset threshold, the system (100) may reset the target VAN approval network for the item corresponding to the empty information to another VAN approval network where payment was successful. Through this reset, the system (100) may correct the conversion rule that is no longer suitable.

[0102] According to one embodiment, the system (100) can respond to changes over time. The issuer's policy may change over time. Even if a specific card was previously unable to make payments through the Fiji approval network, it may subsequently be changed to a supported card. The system (100) can check whether a pre-set validity period has elapsed from the time of update for each item in the conversion mapping table. A validity period may refer to the period during which the item is maintained until it is re-verified. The system (100) can mark items whose validity period has expired as targets for verification by retrying payment through the Fiji approval network. For the marked items, the system (100) can update the payment eligibility by retrying the Fiji approval network during the next payment. Through this re-verification, the system (100) can prevent old conversion rules from remaining unchanged. The system (100) can keep conversion rules up to date even with changes in the issuer's policy.

[0103] According to one embodiment, the system (100) can switch approval networks by considering not only the identification information of the payment card but also the situation information where the payment occurs. Even with the same payment card, the processing performance of each approval network may vary depending on the time of day when the payment occurs or the type of business of the merchant. During specific time periods, payment requests may be concentrated, causing the response delay time of some approval networks to be longer. In specific business sectors, the approval failure rate of a specific approval network may appear relatively high. The system (100) can improve the accuracy of the switching decision by reflecting these situation-dependent characteristics.

[0104] According to one embodiment, the system (100) may collect situational information at the time of a payment request. Contextual information may refer to information including the time of day when the payment occurred, the day of the week, the business classification of the merchant, and the payment amount range. The system (100) may classify the time of payment occurrence into a plurality of pre-defined time zones. For example, the system (100) may classify a day into four time zones: from 0:00 to 6:00, from 6:00 to 12:00, from 12:00 to 18:00, and from 18:00 to 24:00. The system (100) may classify the day of the week into weekdays and weekends, or into seven categories from Monday to Sunday. The system (100) may obtain a business classification code from the merchant information and classify it into any one of the business types including restaurants, convenience stores, gas stations, online shopping, and medical institutions. The system (100) may classify the payment amount into ranges of less than 10,000 won, 10,000 won or more but less than 100,000 won, and 100,000 won or more. The number of time zone divisions, types of business sectors, and boundary values ​​for price ranges are merely examples and are not limited to these; they may vary depending on the settings.

[0105] According to one embodiment, the system (100) may add collected situational information to input items for predicting payment eligibility. The system (100) may input time zones, days of the week, industry classifications, and payment amount ranges together into the classification model in addition to card identification information. The system (100) may convert industry classifications and days of the week, which do not have the meaning of order or size, by placing separate input slots for each type, displaying 1 only in the corresponding slots, and displaying 0 in the remaining slots. The system (100) may convert items that have the meaning of order, such as time zones and amount ranges, into integer values ​​so that the order is maintained, and convert them into ratios so that they become values ​​between 0 and 1. The system (100) may train the classification model using the converted input items to output the payment eligibility probability and the expected response delay time for each approval network together.

[0106] According to one embodiment, the system (100) can select the most suitable approval network for a given situation by combining the payment probability and the expected response delay time. For each approval network, the system (100) can calculate a situational suitability score by subtracting the value obtained by multiplying the normalized expected response delay time by a second weight from the value obtained by multiplying the payment probability by a first weight. The system (100) can select the approval network with the highest situational suitability score. During time periods when payment requests are concentrated, the system (100) can set the second weight for the expected response delay time relatively high so that an approval network with a fast processing speed is selected.

[0107] According to one embodiment, the system (100) can reinforce training data by accumulating payment results by time zone and industry. For each payment transaction, the system (100) can record situation information, a selected approval network, and whether the payment was successful. The system (100) can calculate the success rate for each approval network under the same situation conditions by aggregating the number of successful and unsuccessful payments under the same situation conditions. If the success rate of a specific approval network falls below a preset threshold ratio in a combination of a specific time zone and industry, the system (100) can lower the priority of that approval network under that condition. The system (100) can update the situation-specific prediction criteria by retraining the classification model whenever the accumulated payment results exceed a preset number.

[0109] FIG. 5 conceptually illustrates the overall operation flow of an approval network automatic switching system based on payment card type analysis according to one embodiment.

[0110] Referring to FIG. 5, the system (100) can operate in conjunction with a payment card (505) and a payment terminal (510), and can transmit and receive data for payment approval with a PG approval network (520) and a VAN approval network (530). The configuration and flow shown in FIG. 5 are merely examples and are not limited thereto; some configurations may be omitted, configurations not shown may be added, and the order of some flows may be changed.

[0111] According to one embodiment, when a payment card (505) is contacted, inserted, or tagged with a payment terminal (510), the system (100) may receive card identification information of the payment card (505) from the payment terminal (510). The system (100) may analyze the received card identification information through a card analysis unit (540) and classify the type of the payment card (505). The card analysis unit (540) may extract blank information, an issuer identifier, and a card brand identifier, and classify the type of the payment card (505) by comparing the extracted information with a blank range table.

[0112] According to one embodiment, the system (100) can determine whether payment is possible through the Fiji approval network (520) based on a type classified through the payment possibility determination unit (550). The payment possibility determination unit (550) can compare the classified type and issuer identifier with the list of issuers supported by the Fiji approval network (520). The payment possibility determination unit (550) can transmit a payment approval request corresponding to a payment attempt to the Fiji approval network (520) and receive a response code therefrom. The payment possibility determination unit (550) can determine that payment is possible if the response code is a normal approval code. The payment possibility determination unit (550) can determine that payment is impossible if a rejection response code is received from the Fiji approval network (520) or if a response code is not received within a preset waiting time.

[0113] According to one embodiment, if the system (100) determines that payment through the Fiji approval network (520) is not possible, it can automatically switch the approval network for payment to the VAN approval network (530) through the switching control unit (560). The switching control unit (560) can look up and select the VAN approval network (530) corresponding to the type and issuer identifier of the payment card (505). The switching control unit (560) can terminate the payment session to the Fiji approval network (520) that was in progress and create a payment session with the selected VAN approval network (530). The switching control unit (560) can request payment approval by converting the payment information into the message format of the VAN approval network (530). The switching control unit (560) can complete the payment by receiving a response corresponding to approval from the VAN approval network (530). The switching control unit (560) can automatically switch the approval network without manual intervention by an operator.

[0114] According to one embodiment, the system (100) can accumulate results of payment processing and update classification criteria through a learning and history storage unit (570). The learning and history storage unit (570) can accumulate payment history data, including the type of payment card (505), whether the approval network has switched, and the payment approval result, as learning data. The learning and history storage unit (570) can provide the accumulated learning data to a card analysis unit (540). The card analysis unit (540) can update criteria for classifying the type of payment card (505) and determining whether payment is possible using the provided learning data. Through feedback of these learning results, the system (100) can gradually learn the types of payment cards (505) that cannot be paid in the Fiji approval network (520) and improve the accuracy of the switching determination.

[0115] According to one embodiment, the system (100) can proactively identify new card types by grouping and analyzing common characteristics of cards that have experienced payment failures, going beyond relying on identification information of individual payment cards. For new cards or special cards, it may be difficult for a classification model to accurately determine conversion because there is insufficient training data during the initial launch phase. Even before individual cards are sufficiently trained, the system (100) can reduce gaps in conversion determination by grouping and analyzing failure patterns of cards with similar characteristics.

[0116] According to one embodiment, the system (100) can collect features of payment cards that have been rejected in the Fiji approval network. The system (100) can extract empty information, an issuer identifier, a card brand identifier, and a rejection response code from each rejected payment card. The system (100) can organize the extracted features into a multidimensional feature vector. A feature vector may mean expressing the attributes of each payment card as a sequence of numbers. The system (100) can include attributes that do not have meaning in order or size in the feature vector by converting them by assigning a separate position to each type and displaying 1 only in the corresponding position. The system (100) can separate the top six positions of the empty information into a separate item to reflect similarity at the issuing institution level in the feature vector.

[0117] According to one embodiment, the system (100) can group the configured feature vectors into multiple clusters using a clustering algorithm. Clustering may refer to the process of gathering data with similar features into the same group. When the system (100) uses a k-means clustering algorithm, it can calculate the distance between two feature vectors and group payment cards that are close in distance into the same cluster. When the system (100) uses a density-based clustering algorithm, it can form a cluster in an area where there is a pre-set minimum number of payment cards within a pre-set radius. The system (100) can classify payment cards that do not belong to any cluster as noise and exclude them from cluster formation. The type of clustering algorithm, radius, and minimum number values ​​are merely examples and are not limited thereto, and may vary depending on the settings.

[0118] According to one embodiment, the system (100) can identify a cluster that has grown to a certain size or larger among the formed clusters as a candidate for a new card type. The system (100) can register the cluster as a separate card type if the number of payment cards belonging to a cluster exceeds a preset threshold. For example, the system (100) can register the cluster as a new card type if, within a preset period, the number of different payment cards belonging to a cluster exceeds a specified number (e.g., 30) and the rejection rate of the Fiji approval network of the cluster exceeds a specified level (e.g., 90 percent). For the registered cluster, the system (100) can map the VAN approval network where the payment cards of the cluster have mainly succeeded to the conversion target. The system (100) can calculate which VAN approval network has the highest success rate and designate that VAN approval network as the conversion target for the cluster. The system (100) can reflect this mapping in a conversion mapping table.

[0119] According to one embodiment, the system (100) can determine which cluster a newly presented payment card belongs to and determine a transition. The system (100) can calculate the distance between the feature vector of the newly presented payment card and the center of each cluster. The center of a cluster may refer to the average position of the feature vectors belonging to the cluster. The system (100) can determine the cluster with the shortest calculated distance as the cluster to which the payment card belongs. The system (100) can apply the transition rule of that cluster only if the distance is within a preset threshold distance. If the system (100) determines that the payment card belongs to a cluster with a rejection pattern, it can immediately transition to a van approval network mapped to that cluster even before training data for the individual card is accumulated.

[0120] According to one embodiment, the system (100) can continuously update clusters and conversion rules. The system (100) can update the composition and center of the clusters by re-performing clustering whenever new rejected payment cards accumulate. The system (100) can remove a cluster from the rejection pattern clusters if the rate at which a payment card belonging to a cluster subsequently succeeds in making a payment through the Fiji approval network exceeds a preset standard. Through such updates, the system (100) can exclude card clusters from conversion targets where the rejection pattern has disappeared due to a change in the issuer's policy.

[0121] According to one embodiment, the system (100) can determine in advance whether payment is possible through the Fiji approval network based solely on the card identification information of the payment card before transmitting an actual payment approval request to the Fiji approval network. In normal payment processing, the inability to make a payment can only be recognized after a rejection response code is received following the transmission of a payment approval request to the Fiji approval network. In this case, the time taken to receive the rejection response may lead to a delay in payment processing and may result in the accumulation of unnecessary requests to the Fiji approval network. The system (100) can reduce such delays and unnecessary requests through prior determination.

[0122] According to one embodiment, the system (100) may store a list of non-paymentable cards for preliminary determination in memory (130). The list of non-paymentable cards may refer to a list in which the blank information and issuer identifiers of payment cards confirmed to be non-paymentable through the Fiji approval network are accumulated and recorded. The system (100) may register the blank information and issuer identifiers of payment cards that have previously received a rejection response code from the Fiji approval network into the list of non-paymentable cards. The system (100) may increase the reliability value of an item as rejections are repeated for the same blank information. The reliability value may refer to a numerical value representing the degree of certainty that the item is actually non-paymentable through the Fiji approval network.

[0123] According to one embodiment, when card identification information is obtained from a payment card, the system (100) can check whether the card identification information is registered in the list of cards that cannot be paid before transmitting a payment approval request to the Fiji approval network. If the system (100) finds that the card identification information is registered in the list of cards that cannot be paid and its reliability value exceeds a preset threshold value, the system (100) can switch directly to the VAN approval network without transmitting a payment approval request to the Fiji approval network. The system (100) can reduce the time required for payment processing by omitting the request and response waiting process to the Fiji approval network.

[0124] According to one embodiment, the system (100) may transmit a payment approval request to the Fiji approval network without applying a preliminary judgment to items whose reliability value is below a threshold value. The system (100) may readjust the reliability value of the item based on the response code received as a result. If a rejection response code is received, the system (100) may increase the reliability value, and if a normal approval code is received, it may decrease the reliability value or remove the item from the list of cards that cannot be paid. The system (100) can maintain the accuracy of the preliminary judgment through such dynamic adjustment of the reliability value. The system (100) may update the preliminary judgment criteria even if the payment eligibility changes due to a change in the issuer's policy.

[0125] According to one embodiment, the system (100) can calculate the risk of payment being rejected through the Fiji approval network and determine a transition based on the acquired partial identification information, even when only a part of the card identification information is acquired from the payment card or when the card identification information is damaged and accurate type classification is difficult. During the process of recognizing the payment card at the payment terminal, there may be cases where only a part of the card identification information is acquired due to communication errors or card damage. In this case, the system (100) can select an appropriate approval network based on the risk level without stopping the payment even if complete classification is difficult.

[0126] According to one embodiment, the system (100) can extract partial identification information obtained from a payment card. Partial identification information may refer to information including at least one of a portion of the leading digits of a card number, a portion of an issuer identifier, or a card brand identifier. The system (100) can calculate the Fiji approval network rejection rate for payment cards belonging to empty information intervals that match the extracted partial identification information from payment history data stored in memory (130). The rejection rate may refer to the ratio of the number of rejections to the total number of payment attempts for the payment cards belonging to the corresponding empty information interval.

[0127] According to one embodiment, the system (100) can calculate a risk score based on the calculated rejection rate. The risk score may represent a numerical value indicating the probability that the payment card will be rejected in the Fiji approval network. The system (100) can calculate the risk score such that the higher the rejection rate, the higher the risk score. The system (100) can assign a higher confidence weight to the risk score as the number of digits in the extracted partial identification information increases. The system (100) can apply a conservative correction to the calculated risk score as the amount of partial identification information obtained decreases.

[0128] According to one embodiment, the system (100) can compare a calculated risk score with a preset risk threshold. If the risk score exceeds the risk threshold, the system (100) may skip the payment attempt to the Fiji approval network and switch to the VAN approval network, even if a complete type classification has not been achieved. If the risk score is below the risk threshold, the system (100) may attempt a payment by sending a payment approval request to the Fiji approval network. The system (100) may update the rejection rate of the corresponding empty information interval by reflecting the results of the payment attempt in the payment history data. Through this risk-based switching, the system (100) can reduce payment interruptions and increase the payment success rate even in situations where card identification information is incomplete.

[0130] Artificial intelligence (AI) systems are computer systems that mimic human intelligence; unlike traditional rule-based smart systems, they are equipped with the ability for machines to learn and make decisions autonomously. As the use of these AI systems increases, their cognitive capabilities improve and they gain a more accurate understanding of individual user preferences, leading to the gradual replacement of existing rule-based systems with deep learning-based AI.

[0131] Artificial intelligence technology consists of machine learning and various component technologies based on it. Machine learning includes algorithmic technologies that autonomously analyze and classify the characteristics of input data, while component technologies are techniques that mimic human cognitive and judgment functions by utilizing machine learning algorithms, such as deep learning, and encompass fields such as linguistic understanding, visual understanding, reasoning and prediction, knowledge representation, and motion control.

[0132] Artificial intelligence technology is being applied in various fields; linguistic understanding encompasses the technology of recognizing and processing human language, while visual understanding encompasses the technology of recognizing and processing objects. Reasoning and prediction technologies are used to analyze information and make logical judgments, and knowledge representation technology performs the function of automating human experience into data. Motion control is the technology that manages the movements of autonomous vehicles or robots.

[0133] Generally, applying machine learning algorithms to real-world environments involves a trial-and-error approach. In particular, since deep learning requires numerous iterations and is difficult to execute in actual environments, training is performed through simulations created in a virtual environment on a computer.

[0134] In this invention, artificial intelligence refers to a technology that mimics human learning, reasoning, and perceptual abilities and implements them on a computer, and may include concepts such as machine learning and symbolic logic. Machine learning is an algorithmic technology that independently classifies or learns the characteristics of input data; various artificial intelligence technologies utilize such machine learning algorithms to analyze input data and perform judgments or predictions based on the results.

[0135] Through machine learning, computer software can enhance its data processing capabilities, and neural network models are constructed by modeling the correlations between data. In this process, neural network models extract and analyze features from input data to derive relationships, and optimize parameters through this iterative process. For example, neural network models can learn the relationship between input and output for a given data pair, or learn relationships by identifying regularities solely from the input data.

[0136] Artificial intelligence learning models can be designed to computerize the structure of the human brain, with multiple network nodes simulating the signal transmission of neurons. These nodes are located in layers of different depths to exchange data and can take the form of artificial neural networks, convolutional neural networks (CNNs), and the like. Such models can be trained using methods such as supervised learning, unsupervised learning, and reinforcement learning, and various machine learning algorithms can be applied.

[0137] In particular, CNNs are a type of multilayer perceptron that requires minimal preprocessing and consist of convolutional layers and general neural network layers. CNNs can effectively process 2D input data and demonstrate superior performance in image and speech processing compared to other deep learning architectures. CNNs can be trained using the standard backpropagation method and have the advantage of being easily learned with a small number of parameters.

[0138] Convolutional networks are neural networks composed of nodes that share parameters, and the performance of computer vision tasks has been significantly improved through the combination of large amounts of training data and computational power. As the volume of currently available datasets increases, increasing the size of networks contributes to improving test accuracy. Along with this, the optimal utilization of computing resources has become a critical factor, and distributed implementations of deep neural networks can be used.

[0139] The embodiments described above may be implemented as hardware components, software components, or a combination thereof. For example, the described devices, methods, and components may be implemented using one or more general-purpose or special-purpose computers, such as processors, controllers, arithmetic logic units (ALUs), digital signal processors (DSPs), microcomputers, field programmable gate arrays (FPGAs), programmable logic units (PLUs), microprocessors, etc. The processing unit may execute an operating system (OS) and various software applications running on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to software execution.

[0140] Although embodiments have been described through specific drawings, a person skilled in the art can apply various technical modifications and variations based thereon. For example, appropriate results can be obtained even if the described technology is performed in a different order, or if the components of the described system, structure, device, circuit, etc. are combined in a different form, or if they are replaced with other components or equivalents.

Claims

Claim 1 The automatic approval network switching system based on payment card type analysis includes a memory that stores instructions; The system includes a processor, and the instructions, when executed by the processor, acquire card identification information from a payment card that is contacted, inserted, or tagged at a payment terminal; analyze the acquired card identification information to classify the type of the payment card; determine whether payment is possible through a Payment Gateway (PG) approval network based on the classified type of the payment card; if it is determined that payment is not possible through the Payment Gateway (PG) approval network, automatically switch the approval network for payment to a Value Added Network (VAN) approval network; request payment approval for the payment card through the switched VAN approval network; accumulate payment history data including the type of the payment card, whether the approval network was switched, and the payment approval result as training data; update the classification of the payment card type and the determination of switching the approval network based on the accumulated training data; extract Bank Identification Number (BIN) information corresponding to the leading digit of the card number, the issuer identifier, the card brand identifier, and the rejection response code from each payment card whose payment was rejected by the Payment Gateway (PG) approval network; and each of the extracted features Constructing a multidimensional feature vector in which the attributes of a payment card are expressed as a sequence of numerical values, wherein attributes that do not have meaning in order or size are converted by assigning a separate position to each type and marking 1 only in the corresponding position, and included in the said feature vector; separating the top six positions among the said empty information into a separate item to reflect similarity at the issuing institution level in the said feature vector; and grouping the said constructed feature vectors into multiple clusters using a clustering algorithm corresponding to the process of grouping data with similar characteristics into the same group, andIf the number of payment cards belonging to a certain cluster exceeds a preset threshold and the rejection rate of the Fiji approval network for that cluster exceeds a preset level, the cluster is registered as a separate card type as a cluster with a rejection pattern; the success rate of which VAN approval network has the highest among the VAN approval networks where payment cards belonging to the registered cluster have successfully made payments is calculated, and that VAN approval network is designated as the target for conversion of the cluster; the distance between the feature vector of the newly presented payment card and the center of each cluster corresponding to the average position of the feature vectors belonging to each cluster is calculated, and the cluster with the shortest calculated distance is identified as the cluster to which the newly presented payment card belongs, provided that the conversion rule for that cluster is applied only if the distance is within a preset threshold distance; if the newly presented payment card is determined to belong to the cluster with the rejection pattern, it is immediately converted to the VAN approval network mapped to that cluster even before training data for individual cards is accumulated; whenever a new rejected payment card is accumulated, the clustering is performed again to update the cluster composition and centers, and in which cluster If the rate at which a payment card belonging to the cluster succeeds in making a payment through the Fiji approval network subsequently exceeds a preset standard, the cluster is removed from the cluster having the aforementioned rejection pattern, and a list of payment-unavailable cards is stored by accumulating and recording the blank information and issuer identifiers of payment cards confirmed to be unavailable for payment through the Fiji approval network as items, wherein as rejection is repeated for the same blank information, a confidence value corresponding to a numerical value of the degree of certainty that the item is actually unavailable for payment in the Fiji approval network is increased, and when card identification information is obtained from the payment card, prior to transmitting a payment approval request to the Fiji approval network, it is checked whether the card identification information is registered in the aforementioned payment-unavailable card list.A system that, when the relevant card identification information is registered in the aforementioned list of cards ineligible for payment and its reliability value exceeds a preset threshold value, does not transmit a payment authorization request to the Fiji authorization network but immediately switches to the VAN authorization network, and for items where the reliability value is below the threshold value, transmits a payment authorization request to the Fiji authorization network, and controls the system to increase the reliability value if a rejection response code is received as a result, and to decrease the reliability value or remove the item from the aforementioned list of cards ineligible for payment if a normal authorization code is received. Claim 2 In claim 1, when the instructions are executed by the processor, the system obtains the Bank Identification Number (BIN) information corresponding to the leading digit of the card number, the issuer identifier, and the card brand identifier from the payment card as card identification information; classifies the payment card into one of the types of credit card, debit card, prepaid card, corporate card, and special purpose card by comparing the obtained BIN information with a pre-stored BIN range table; and, in order to determine whether payment is possible through the Fiji Authorization Network, first checks whether the payment card is included in the list by comparing the classified type and the issuer identifier with the list of issuers supported by the Fiji Authorization Network; if the payment card is included in the list, transmits a payment authorization request to the Fiji Authorization Network and checks the received response code; if the response code is a pre-defined normal authorization code, it is determined that payment through the Fiji Authorization Network is possible; and if the response code is a rejection response code corresponding to issuer non-support, card type non-support, or transaction rejection, it is determined that payment through the Fiji Authorization Network is not possible. Even if the payment card is not included in the above list or a response code is not received from the Fiji approval network within a preset waiting time, it is determined that payment through the Fiji approval network is not possible; if it is determined that payment through the Fiji approval network is not possible, a VAN approval network corresponding to the type and issuer identifier of the above payment card is retrieved and selected from the pre-stored network correspondence information, the payment session to the Fiji approval network that was in progress is terminated, and a payment session with the selected VAN approval network is created, and the approval network is automatically switched by converting the payment information obtained from the above payment card into the message format of the VAN approval network and requesting payment approval.If only a portion of the card identification information is obtained from the payment card or if the card identification information is damaged and the type classification of the payment card is not performed, partial identification information is extracted from the information obtained from the payment card, including at least one of a portion of the leading digits of the card number, a portion of the issuer identifier, or a card brand identifier; from the payment history data stored in the memory, for payment cards belonging to an empty information interval that matches the extracted partial identification information, a Fiji approval network rejection rate is calculated corresponding to the ratio of the number of rejections to the total number of payment attempts for the payment cards belonging to the said empty information interval; based on the calculated rejection rate, a risk score is calculated corresponding to a value that quantifies the probability that the payment of the payment card will be rejected by the Fiji approval network; the calculated risk score is compared with a preset risk threshold, and if the risk score exceeds the risk threshold, the payment attempt to the Fiji approval network is omitted and switched to the VAN approval network even if the type classification of the payment card has not been performed, and if the risk score is below the risk threshold, a payment approval request is transmitted to the Fiji approval network to attempt payment; and A system that controls updating the rejection rate of the corresponding empty information interval by reflecting the result of the payment attempt in the payment history data. Claim 3 In claim 1, when the instructions are executed by the processor, the system constructs training data for classifying the type of the payment card, wherein the card identification information includes BIN information, an issuer identifier, a card brand identifier, and a payment request amount as input items, and the answer value is whether the payment card was eligible for payment through the Fiji authorization network; regarding the constructed training data, items for which the order does not have meaning, such as BIN information and the issuer identifier, are converted by assigning a separate input position for each type of item, displaying 1 only in the corresponding position and 0 in the remaining positions, and the payment request amount is converted by a ratio such that the value between a preset minimum and maximum value becomes a value between 0 and 1; and using the preprocessed training data, a classification model is trained to output a payment possibility probability through the Fiji authorization network from the input card identification information as a value between 0 and 1, wherein the difference between the payment possibility probability output by the classification model and the answer value indicating actual payment possibility is calculated as an error, and the classification model is trained by iteratively adjusting the internal weights of the classification model in a direction in which the calculated error is reduced. A system for training, and when the probability of payment possible through the Fiji approval network calculated by inputting card identification information of a newly contacted, inserted, or tagged payment card into the trained classification model is less than a preset threshold probability, automatically switching the approval network for payment to a VAN approval network, and retraining the classification model by adding training data in which the payment approval result of the VAN approval network resulting from the automatic switching is the correct answer value to the training data, thereby controlling the gradual learning of the types of payment cards that cannot be paid in the Fiji approval network. Claim 4 delete Claim 5 delete

Citation Information

Patent Citations

  • Business routing method and apparatus

    CN106878169A

  • Credit card Intelligent payment systems and payment processing method

    KR1020180041389A

  • Electronic device, method, and non-transitory computer readable storage medium for performing payment

    KR1020250015605A

  • Methods and systems for product identification and computer routing services

    US20170083877A1