KR20160017629A - Predictive fraud screening - Google Patents

Abstract

The present invention relates to methods, systems and computer program products for blocking transactions. Transactions stored in the transaction database can provide a statistical basis for determining the probability of fraud for pending transactions. The fraud cost can be determined for each of a plurality of possible actions based on the fraud probability, and the action provides the lowest fraud cost selected. Cumulative fraud costs can be determined for a set of transactions in the database. If taking action at the lowest fraud cost for a transaction would result in a higher cumulative fraud cost, an action with a higher fraud cost for the transaction could be selected. Transactions can be added to the database, and the probability of fraud is adjusted based on the amount of time since the transaction was approved. The statistical basis can be optimized based on the cost error.

Description

Predictive thinking prevention {PREDICTIVE FRAUD SCREENING}

The present invention relates generally to computers and computer systems, and more particularly to methods, systems, and computer program products for screening transactions against fraud.

In the travel industry, airline tickets are often sold through indirect sellers such as travel agencies. Indirect sellers will generally check for available flights or other travel services that meet traveler's travel plans, and once services that match are found, book a service for travelers, We collect payment. Charges are often collected by charging the cost of travel services purchased to a credit card account provided by the traveler, and the indirect seller or validating carrier acts as the merchant.

Credit card transactions typically involve a two-stage process of authorization and settlement. At the time of the transaction, transaction information such as the purchase amount, the merchant's identity, the credit card account number, and the expiration date is transmitted from the merchant to the issuing bank. The issuing bank may then check the account to verify that the credit card is valid and that the credit line is sufficient to allow the transaction. If the bank approves the transaction, the merchant completes the transaction and issues a ticket to the traveler. To receive a payment, the merchant may send a batch of approved certifications to the "acquiring bank" at the close of the business day (business day). The acceptance bank can then reconcile and transmit certificates to issuing banks, typically via a card network or clearing house, deposit funds in the merchant's account do. The reserves are then transferred from the issuing bank to the accepting bank and a bill is sent by the issuing bank to the cardholder.

Unfortunately, credit cards are often used by unscrupulous individuals who make unauthorized purchases using improperly acquired or stolen credit cards to illegally purchase tickets. When a real cardholder notices an unauthorized purchase, the cardholder may dispute the charge with the issuing bank. These disputes typically result in a "chargeback" to the merchant for the cost of the transaction. Chargebacks may be received up to a maximum number of months for which travel services are typically used after a transaction occurs. Thus, fraudulent credit card transactions generally cause significant damage to merchants who can not recover the cost of travel services.

Thus, an improved system for analyzing transactions and detecting fraud to reduce the incidence of illegal charges and reduce losses incurred to merchants and travel service providers due to illegal purchases of travel services Methods, and computer program products are needed.

In one embodiment of the invention, a method is provided for blocking transactions. The method includes receiving first data characterizing a first transaction and determining a first probability of a first transaction based on the first data. The method includes receiving second data characterizing a second transaction occurring during a period during which a chargeback for a first transaction can be received and receiving second data characterizing a first probability and a second transaction based on the amount of time since approval of the first transaction And determining a second probability of a first transaction to be fraudulent. The method may then determine a third probability of a second transaction based on the second data and a second probability based at least in part.

In another embodiment of the invention, an apparatus is provided for blocking transactions. The apparatus includes a processor and a memory coupled to the processor. The memory, when executed by the processor, includes instructions to cause the device to receive first data characterizing a first transaction and to determine a first probability of a first transaction based on the first data. The device may also receive second data characterizing a second transaction occurring during a period during which a chargeback for the first transaction can be received and may include a first probability and an amount of time after the first transaction has been approved The first transaction can determine the second probability of fraud. The device may also determine a third probability of a second transaction based on the second data and a second probability at least partially.

In another embodiment of the present invention, a computer program product is provided that includes a non-transitory computer-readable storage medium including instructions. The instructions, when executed by the processor, may be configured to cause the processor to receive first data characterizing the first transaction and to determine a first probability of a first transaction based on the first data. The instructions may also cause the processor to receive second data characterizing a second transaction occurring during a period during which a chargeback for the first transaction can be received, and after the first probability and after the first transaction has been approved The first transaction may determine the second probability of fraud based on the amount of time of the first transaction. The processor may also determine a third probability of a second transaction based on the second data and a second probability at least partially.

In another embodiment of the present invention, a method is provided for blocking transactions involving receiving data defining a first transaction. The method may further comprise determining a first fraud cost for taking a first action and a second fraud cost for taking a second action. In response to the second fraud cost exceeding the first fraud cost, the method may determine a reduction in the cumulative fraud cost for a testing set of transactions that may be caused by taking a second action. The method may also determine a first difference between the first fraud cost and the second fraud cost and take a second action if the decrease is greater than the first difference.

In another embodiment of the invention, an apparatus is provided for blocking transactions. The apparatus includes a processor and a memory coupled to the processor. The memory, when executed by the processor, includes instructions for causing the device to receive data defining a first transaction. The instructions may also be configured to cause the device to determine a first fraud cost for taking the first action and a second fraud cost for taking the second action. In response to the second fraud cost exceeding the first fraud cost, the device may determine a reduction in the cumulative fraud cost for the testing set of transactions by taking a second action, The first difference can be determined. If the decrease is greater than the first difference, then the device can take a second action.

In another embodiment of the present invention, a computer program product is provided that includes a non-transient computer-readable storage medium including instructions. The instructions, when executed by the processor, may be configured to cause the processor to receive data defining a first transaction. The instructions may also be configured to determine a first fraud cost for taking a first action and a second fraud cost for taking a second action. In response to the second fraud cost exceeding the first fraud cost, the instructions can determine a decrease in the cumulative fraud cost for the testing set of transactions by taking the second action, and the difference between the first fraud cost and the second fraud cost The first difference can be determined. If the decrease is greater than the first difference, the instructions may take a second action.

BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate various embodiments of the invention and, together with the general description of the invention given above and the detailed description of the embodiments given below, .
1 is a schematic diagram of an exemplary operating environment including a plurality of computing systems communicating over a network.
Figure 2 is a schematic diagram of the exemplary computing system of Figure 1;
3 is a schematic diagram of a transaction blocking system including a fraud screening module, a fraud probability module, and a transaction database.
Figure 4 is a graphical depiction of the relationship between the fraud cost and the fraud probability for a transaction blocked by the transaction blocking system of Figure 3;
5 is a graphical representation of a probability density function that illustrates the relationship between the probability of receiving a chargeback for a transaction and the amount of time after the acceptance of a transaction.
Figure 6 is a flow diagram of a transaction cut-off process that may be performed by the transaction cut-off system of Figure 3;

Embodiments of the present invention are directed to methods and systems for determining whether to approve or deny a transaction, such as a transaction for purchasing a travel service, by determining a fraud cost for the transaction. Embodiments of the invention may be implemented by a transaction blocking system that includes one or more networked computers or servers. Networked computers may include a Global Distribution System (GDS) and may provide processing and database functions for travel-related systems and modules that analyze transactions against fraud .

The transaction blocking system may include a transaction database that includes a set of historical transactions that provide a statistical basis for estimating fraud probabilities for future transactions. To initialize the transaction database, an initial fraud probability may be set for each of the transactions in the set of historical transactions. At appropriate intervals, or as new information is received, the fraud probability can be updated. The new information may include the expiration of an acceptable chargeback period without receiving a chargeback or receiving a chargeback for a transaction in the transaction database. The probability of fraud estimates can now be maintained with the latest information available thereby.

As requests for new transactions are received, the probability of fraud can be estimated for new transactions based on data in the transaction database. The fraud probability can be parameterized based on a learning algorithm that uses an error function to capture additional fraud costs caused by erroneous predictions. The fraud cost can thereby be associated with making an optimal decision based on both incorrect predictions and correct predictions. Since the results of the authorization transactions are constantly known, once the results are known, the fraud estimates for the authorization transactions are assigned a probability of fraud of zero (no chargeback received) or 1 (unity) Can receive.

Embodiments of the present invention may further include features that allow the system to learn by approving transactions to be rejected based only on the fraud cost for the transaction. For this, a predicted change in the cost of fraud can be determined across multiple transactions from information obtained by allowing transactions that would otherwise have been rejected. Transactions can be allowed if the expected improvement in the ability to estimate the fraud costs for other transactions offsets the fraud costs to allow the transaction. Embodiments of the present invention may also include features that extend on-line fraud screening by taking into account probability information about previous transactions as well as information relating to transactions for which authorization decisions should be made.

1, an operating environment 10 in accordance with an embodiment of the present invention may include one or more travel service provider systems, such as a Global Distribution System (GDS) 12, a carrier system 14, One or more indirect merchant systems, such as a travel agent system 16, a receiving bank system 18, an issuing bank system 20 and a billing and settlement plan (BSP) system 22 . Each of the GDS 12, the airline system 14, the travel agent system 16, the acceptance bank system 18, the issue bank system 20 and the BSP system 22 may communicate via the network 24. The airline system 14 may include a computer reservation system (CRS) or billing system that allows the GDS 12 or the travel agent system 16 to book or pay for air tickets. The airline system 14 may also communicate with other airline systems (not shown), either directly or via the GDS 12, to enable a valid airline to sell tickets for seats provided by an operating carrier. Can interact. The operating carrier may then charge the airline available for the services provided. Charges between indirect vendors and travel service providers may be provided by the BSP system 22. [ Revenue can also be redeemed directly from indirect sellers (i.e., without the use of a BSP), in which case the travel provider (e.g., a valid airline) will ensure that any issued memos are paid . The network 24 may include one or more private or public networks (e.g., the Internet) that may enable exchange of data.

The GDS 12 may be used by the travel agencies, valid airlines or other indirect sellers to schedule reservations on the airline system 14 via the GDS 12 to communicate between the airline system 14 and the travel agent system 16 As shown in FIG. The GDS 12 may maintain links to a plurality of airline systems via the network 24, which may allow the GDS 12 to route reservation requests from a valid airline or travel agent to a corresponding operating airline. The airline system 14 and the travel agent system 16 may thereby schedule flights on multiple airlines through a single connection to the GDS 12.

The travel agent system 16, the acceptance bank system 18 and the issue bank system 20 may be configured to exchange the data necessary to execute the transaction. To this end, at the time of transaction, the travel agent system 16 may send an authentication request to the issuing bank system 20. In response to receiving the authentication request, the issuing bank system 20 may verify that the credit card account is valid and that the account has sufficient residual credit to cover the transaction volume. The issuing bank system 20 may then send an authentication response to the travel agent system 16. Such a response may indicate that the transaction has been approved, rejected, or may request additional information. Once the transaction is complete, the travel agent system 16 may send data to the acceptance bank system 18 that characterizes the transaction. This data may be transmitted as part of a batch file at the end of the same period as at the end of the business day. The receiving bank system 18 can then deposit the reserves into the account of the travel agent and retrieve the deposits from the corresponding issuing banks of the credit cards used to purchase the travel services.

The BSP system 22 may be configured to receive data from a travel agent reporting a ticket sale in the name of the operating airline or from a ticketing office of a valid travel agent. In the United States, Airline Reporting Corporation (ARC) typically provides such services. In any case, the BSP can act as a business process outsourcer (BPO) that provides a set of clearing houses to determine the accounts between travel agents and valid airlines. Other systems (not shown) are also connected to the network 24 to determine the accounts between operational and valid airlines, such as systems operated by the IATA Clearing House (ICH) or Airlines Clearing House (ACH) . In any event, these various clearinghouse systems may facilitate collection of fares by the operating airline to provide services sold by other business entities.

2, the GDS 12, the airline system 14, the travel agent system 16, the acceptance bank system 18, the issue bank system 20 and the BSP system 22 of the operational environment 10 And may be embodied on one or more computer devices or systems, such as the exemplary computer system 26. The computer system 26 includes a processor 28, a memory 30, a mass storage memory device 32, an input / output (I / O) interface 34 and a Human Machine Interface ) ≪ / RTI > The computer system 26 may also be operatively coupled to one or more external resources 38 via the network 24 or the I / O interface 34. External resources may include servers, databases, mass storage devices, peripheral devices, cloud-based network services, or any other suitable computer resource that may be used by computer system 26 But are not limited to these.

The processor 28 may be a microprocessor, microcontrollers, digital signal processors, microcomputers, central processing units, field programmable gate arrays, programmable logic devices, state machines, Includes one or more devices selected from logic circuits, analog circuits, digital circuits, or any other device that manipulates signals (analog or digital) based on operational instructions stored in memory 30 can do. The memory 30 may be a read-only memory (ROM), a random access memory (RAM), a volatile memory, a non-volatile memory, a static random access memory (SRAM) A dynamic random access memory (DRAM), a flash memory, a cache memory, or any other device capable of storing information. The mass storage memory device 32 may be a hard drive, an optical drive, a tape drive, a non-volatile solid state device, or any other device capable of storing information, Devices.

The processor 28 may operate under the control of the operating system 40 residing in the memory 30. [ The operating system 40 may be configured to manage computer resources so that computer program code embodied as one or more computer software applications such as applications 42 residing in memory 30 may have instructions to be executed by the processor 28. [ can do. In an alternative embodiment, the processor 28 may execute the application 42 directly, in which case the operating system 40 may be omitted. The one or more data structures 44 may also reside in the memory 30 and may be used by the processor 28, the operating system 40 or the application 42 to store or manipulate data.

The I / O interface 34 may provide a machine interface that operatively connects the processor 28 to other devices and systems, such as the network 24 or external resources 38. The application 42 may thereby communicate by way of the I / O interface 34 to provide various features, functions, applications, processes or modules, including embodiments of the invention, ) Or an external resource (38). The application 42 may also have program code that is executed by one or more external resources 38 or may include functions or signals provided by other systems or network components external to the computer system 26 Lt; / RTI > In addition, if endless hardware and software configurations are possible, those skilled in the art will appreciate that embodiments of the present invention may be implemented on a computer system 26 that is external to the computer system 26 and distributed among multiple computers or other external resources 38 , Applications provided by computing resources (hardware and software) provided as a service through the network 24, such as a cloud computing service.

The HMI 36 may be operatively coupled to the processor 28 of the computer 26 in a manner known to allow a user to interact directly with the computer 26. [ The HMI 36 may include video or alphanumeric displays, touch screens, speakers, and any other suitable audio and visual indicators capable of providing the user with the data. The HMI 36 also includes an alphanumeric keyboard, pointing device, keypads, and / or keypad, which can accept commands or input from a user and send the entered input to the processor 28. [ Input devices such as pushbuttons, control knobs, microphones, and the like, and controls.

The database 46 may reside on the mass storage memory device 32 and may be used to collect and organize the data used by the various systems and modules described herein. The database 46 may include support data structures for storing and organizing data and data. In particular, the database 46 may be arranged in any database organization or structure including, but not limited to, an associated database, a hierarchical database, a network database, or combinations thereof. A database management system in the form of a computer software application executing as instructions on the processor 28 may be used to access information or data stored in records of the database 46 in response to a query, May be dynamically determined and executed by the operating system 40, other applications 42, or one or more modules. In one embodiment of the present invention, the database 46 may include a transaction database 56 (FIG. 3) that includes historical transaction data that provides a statistical basis for estimating the probability of fraud for pending transactions .

3, the transaction blocking system 50 may include a fraud blocking module 52, a fraud probability module 54, and a transaction database 56. [ The transaction blocking system 50 includes a GDS 12, an airline system 14, a travel agent system 16, a receiving bank system 18, an issuing bank system 20, a BSP system 22, Lt; RTI ID = 0.0 > and / or < / RTI > In operation, the fraud blocking module 52 may receive transaction data 58 that characterizes the pending transaction. The transaction data may include a transaction authorization request and may include a plurality of parameters characterizing the transaction. These parameters include the identity of the purchaser, the price charged for the travel service, the payment method, the account being withdrawn for payment (e.g., credit card number and issuing bank), the name or company associated with the account, , The name of the passenger being arranged on the ticket, the origin of the flight, the destination of the flight, the time before the start of the flight, the travel date and time, one or more stopover locations, the class of service, The type of travel goods being sold, the number of travelers to whom tickets are to be issued, the identity of the airline providing the service, the identity of the merchant or merchant, or the location of the transaction, transaction, or IP of the requesting device Any other suitable data characterizing the address, or any other suitable data characterizing the transaction, But are not limited thereto.

The fraud blocking module 52 may analyze the transaction data 58 using the estimated fraud probability P _E (e.g., the estimated probability of receiving a chargeback) and statistical data. The estimated fraud probability P _E can be determined by the fraud probability module 54 that can generate the estimated fraud probability P _E based on the fraud probability function. The parameters of the fraud probability function may include a set of parameters that are determined based on data in the transaction database 56. [ Transaction data 58 may also be used to adjust parameters in fraud blocking module 52 and fraud probability module 54 and to update transaction database 56 based on prediction errors.

Referring now to FIG. 4, graph 60 includes a horizontal axis 62 corresponding to the transaction probability P, and a vertical axis 64 corresponding to the anticipated fraud costs to approve the transaction. Those skilled in the art will appreciate that the scale of the horizontal axis 62 and the vertical axis 64 of the graph 60 may be distorted to more clearly illustrate embodiments of the present invention. The graph 60 includes four functions or curves 66a, 66b, 68, The functions represented by the curves 66a, 66b, 68 and 70 may be used to calculate the fraud costs for different actions taken in response to receiving the transaction and for different assumptions about the nature of the transaction Can be output. According to one embodiment of the present invention, an exemplary function for minimizing fraud costs for a transaction based on the probability of fraud P is to use portions of curves 66a, 66b, 68, 70 as described below . ≪ / RTI >

In an exemplary embodiment, the fraud cost C _F for approving all transactions can be represented by curve 66a. To this end, the curve 66a may include a line corresponding to the merchant liability L times the fraud probability P, and thus C _F = L x P. That is, curve 66a may be defined by a line crossing the vertical axis at zero and having a slope = L, and may represent a fraud cost versus fraud probability P for acknowledging all transactions without performing any additional actions . In some cases, the merchant's liability L may be equal to the trading volume A. The merchant's liability for a given transaction may also be determined by comparing the estimated costs of processing the chargeback, generation and management of the Agency Debit Memo (ADM) and any other additional charges incurred from authorizing fraudulent transactions Costs. Thus, curve 66a may reflect anticipated fraud costs in addition to trading volume A. [

The fraud costs for transactions for which a security check is performed before accepting or rejecting a transaction can be represented by curve (68). Security checks may include, for example, requesting further identification from a traveler, such as an email, a cordless phone, or a code sent to another account associated with the cardholder. The transaction may be approved if the security check passes, and may be rejected if the security check is not passed. The security check may also include performing a manual review of the transaction or the execution of an additional security layer. In any case, performing a security check can add additional costs to the transaction. Exemplary costs include fixed costs such as costs required for implementation (e.g., purchases of computer systems, data connections, terminals, etc.), and cost per transaction (e.g., Costs incurred by the providing company), and costs associated with loss of legitimate sales due to a traveler failing to complete a transaction in response to a security check.

These costs may be reflected in the curve 68 intersecting the vertical axis 64 at the fiducial marker 72. [ For a transaction with a probability of fraud of P = 0%, the cost of fraud at the fiducial marker 72 is less than the cost of fraud, since the expected chargeback cost or the expected chargeback amount is $ 0.00 (i.e., the probability of fraud P times the chargeback amount) (For example, $ 1.50 per transaction) to perform a gross inspection. As the fraud probability P to receive a chargeback increases, the fraud costs for each transaction, including the security checks, are reduced due to security checks identifying fraudulent charges, as indicated by the downward slope of curve 68 can do. That is, rejecting fraudulent transactions identified by the security check may reduce the number of chargebacks, thereby offsetting the security check costs. Thus, the curve 68 can be tilted downward as the fraud probability P increases and higher percentage transactions are rejected.

In some cases, the security check may have the probability of returning a false negative. That is, security checks can sometimes provide a false indication that a transaction is legitimate when the transaction is actually fraudulent. The costs associated with false negative results may tend to increase the anticipated fraud costs to approve transactions as the probability of fraud P increases. Thus, in some embodiments, the slope of curve 68 may be even or even positive due to the number of fraudulent transactions made through security checks. Thus, those skilled in the art will appreciate that embodiments of the present invention are not limited to the exemplary curve 68 shown in FIG.

The fraud costs that reject all transactions can be represented by curve 70 and represent the costs associated with lost sales by rejecting non-fraudulent transactions. Since the revenue is not lost by rejecting the fraudulent transaction, the curve 70 can provide a zero fraud cost for transactions with a probability of fraud P = 100%. The curve 70 may thus include a line intersecting the horizontal axis 62 at the fiducial marker 74 and may be defined by C _F = A x (1-P) Cost C _F = A has a fraud probability of P = 0%.

To minimize the cost of fraud, embodiments of the present invention determine whether to approve the transaction, whether to request a security check and reject the transaction based on it, or reject the transaction based on the fraud probability P . As shown in the exemplary embodiment depicted by Figure 4, curve 66a intersects curve 68 at reference marker 76a (e.g., a probability of fraud of P = 2%), and curve 68 Crosses the curve 70 at the fiducial marker 78 (e.g., P = 92% fraud probability). Thus, in the illustrated embodiment, for fraud probability P between 0% and 2%, the lowest fraud cost is provided by curve 66a. To minimize the fraud costs for this exemplary embodiment: (1) transactions with a fraud probability P between 0% and 2% should be approved without requesting a security check; (2) Transactions with fraud probability P between 2% and 92% should be subjected to the requested security checks and transactions are either approved or rejected based on the results; Transactions with fraud probability P greater than 92% should be rejected without performing a security check. Those skilled in the art will appreciate that both the number and composition of the illustrated curves 66a, 66b, 68, 70 are for illustrative purposes only. Embodiments of the present invention are thus not limited to exemplary curves 66a, 66b, 68, 70, fiducial markers 72, 74, 76a, 76b, 78 or the ranges and values shown in Fig.

Credit card transactions may include "card presence" and "card absent" transactions. Card non-existent transactions occur when a buyer is physically absent, for example, when a transaction is carried out over the telephone or through an online travel agent website. Since the buyer does not carry a card during non-card transactions, the indirect seller may be more difficult to verify that the buyer is the actual cardholder. In some cases, the billing agreements between indirect sellers and airlines may allow liability for fraudulent charges to be transferred to indirect sellers if the underlying transaction is a card-less transaction.

Thus, for tickets sold in the indirect marketplace where the airline is a merchant, liability for fraudulent sales may ultimately be the responsibility of the indirect seller. Under these scenarios, in response to receiving the chargeback, the merchant carrier may send the ADM to the indirect merchant requesting the indirect seller to pay the amount of the fee plus ADM processing fee. Indirect sellers may be required to reimburse the airline for charges in the ADM if the indirect seller can not show that the sale was a "card present" sale. In cases where the transaction in question has been done through an online travel agent's website, the transaction will typically be a non-card transaction and the indirect seller may not be able to prove the sale. Thus, when airlines decide whether to approve or reject a transaction, or to ask for more information about a transaction, they may want to consider this difference in accountability between indirect sales and direct sales.

To this end, airlines can determine the percentage of total transactions that are online transactions for each indirect seller. The airline can then adjust the merchant's liability to fraud for each indirect seller based on the percentage of their total transactions, which are online transactions. In the exemplary graph 60, the travel agent in question can perform 90% of their transactions through the online website and 10% of their transactions with the person in the office. In this example, the airline may want to adjust the trader's liability L for each transaction made by this travel agent to 90%. This adjustment may be based on the expectation that losses will be restored to a large percentage of fraudulent transactions carried out online by this travel agency. The downward adjustment at the merchant's liability L may cause a corresponding decrease in the slope of the line corresponding to the fraud cost C _F , as indicated by curve 66b. Curve 66b may in turn generate an intersection with curve 68 that moves from fiducial marker 76a to fiducial marker 76b.

Therefore, if the airline considers the amount of online transactions performed by the travel agency when minimizing the fraud costs: (1) transactions with a fraud probability P of 0 to 20% must be approved without asking for a security check; (2) Transactions with fraud probability P of 20% to 92% shall be requested for security checks and transactions shall be approved or rejected based on the results; And (3) transactions with fraud probability P greater than 92% should be rejected without performing a security check. For purposes of clarity, the slope of curve 68 has not been adjusted in the above example to account for a reduction in merchant liability L. However, those skilled in the art will appreciate that changes in the trader's liability L may also affect the positions of the points defining the curve 68 and additionally contribute to the movements at the reference points 76b and 78 I will understand.

The above selection process may depend on the assumption that the fraud probability P is correct. However, the fraud probability P for determining which action to take may be the estimated fraud probability P _E provided by the fraud probability module 54 and may be different from the actual fraud probability P _A that the chargeback will be received for the transaction have. In order to reduce this prediction error and adapt the system to changing conditions, embodiments of the present invention can determine the "cost error" C _E resulting from predicting additional costs to the merchant, or "false & And may adjust parameters in fraud blocking module 52, fraud probability module 54, or transaction database 56 to minimize cost error C _E.

The transaction blocking system 50 uses the historical data in the transaction database 56 to compare the estimated probabilities P _E for approved transactions with known probabilities based on whether a chargeback has been received for the transaction in question, C _E can be determined. As an example, the actual fraud probability P _A for an approved transaction to be analyzed may be represented by a fiducial marker 74 at 100% on the horizontal axis. The actual fraud probability P _A can be set to 100% since the chargeback was received regarding the transaction in question. The estimated fraud probability P _E generated by the fraud engine for the transaction in question can be represented by a fiducial marker 80 located at about 90% on the horizontal axis 62. Thus, the prediction error for this transaction may be about 10%.

Based on the actual fraud probability P _A , the optimal decision to minimize E _COF was provided by curve 70, so if the fraud probability P _E is correct, the transaction would have been rejected without requesting a security check. Conversely, based on the estimated fraud probability P _E generated by the fraud probability module 54, a determination was made by curve 68 that the fraud blocking module requested a security check. In this example, the cost error C _E caused by the inaccuracies in the estimated fraud probability P _E is the fraud cost C provided by the curve 68 for the actual fraud probability P _A represented by the fiducial marker 74 _F and the fraud cost C _F provided by the curve 70, for example, about $ 0.50. That is, in this particular example, the estimated cost of choosing to conduct a security check rather than simply denying the transaction would add about $ 0.50 to the fraud cost C _F for the transaction in question. In an embodiment of the present invention, the cost error C _E for a plurality of historical trades can be minimized. That is, the transaction blocking system 50 may adjust the parameters of the fraud blocking module 52, the fraud probability module 54, or the transaction database 56 to minimize the cost error C _E for the set of historical transactions.

To parameterize the algorithm, historical data stored in the transaction database 56 may be used to minimize the sum of the error functions. Methods that can be used to ensure correct fit and convergence of error minimization functions include randomly initializing a set of parameters [theta], and a set of parameters using a stochastic gradient descent lt; / RTI > to < RTI ID = 0.0 > θ. < / RTI > This may include the use of the regularization parameter l. To this end, the history data may be stored in a training set TR (e.g., 60% of historical data); A cross-validation set CV (e.g., 20% of historical data), and a testing set TE (e.g., 20% of historical data). The regularization parameter lambda can be used to avoid exceeding or not exceeding fit while minimizing the following function:

Determining the error includes: (1) using the training set TE to determine the values of the set of parameters? As a function of?; (2) using a cross-check set CV to distinguish between different values of λ so that J (TR) to J (CV) are minimized and to select λ; And (3) selecting a set of parameters &thetas; corresponding to the determined lambda. The "representative" loss can then be determined by J (TE).

Referring now to FIG. 5, graph 90 shows a plot 92 that can represent an exemplary probability density f (t) for a time to receive a chargeback for a transaction, or a charge rejection density function . The probability density f (t) may be determined empirically based on the historical transaction data in the transaction database 56. [ For this, the probability density f (t) may be based on a reference distribution determined from the amounts of time between the approval date for each transaction in the set of transactions in transaction database 56 and the date of receipt of the chargeback.

The fraud probability P (t) from which the charge rejection will be received by time T from the approval date of the transaction can then be determined based on the area under the plot 92, as given by:

The fraud probability P may be determined based on the set of transactions in the transaction database past the chargeback period. That is, for transactions outside the chargeback period, the transaction is denied payment, and thus it can be known whether it is fraudulent. The fraud probability P for a set of grant transactions over a chargeback period can thus be determined based on the ratio between the total number of transactions in the set and the number of chargebacks in the set. The total area under the plot 92 may be one, as shown below:

Where T _max is the time limit or chargeback period for receiving a chargeback of the transaction. That is, once t> T _max , the charge rejection may no longer be received, and hence the probability density f (t) = 0 out of T _max .

For transactions that have not received a chargeback but are still within their chargeback period, the probability P (t) for t can be determined based on the past amount of time since the approval of the transaction as follows:

As can be seen from the equation for P (t) and the exemplary plot 92, the fraud probability P (t) for a given transaction can drop over time without receiving a chargeback. The expected amount of chargeback can be determined for the transaction by multiplying the cost of the chargeback (e.g., transaction cost) times the fraud probability P (t).

To initialize the transaction database 56, an initial fraud probability P _I may be established for all transactions. This value can be a predetermined value, or it can be based on transactions for which a previous fraud blocking strategy such as the previous accounting period has been applied. The initial fraud probability P _I can be estimated based on the authorization decisions applied to the history transactions. On average, an observation or statistical fraud rate for authorization transactions can be measured to be a specific percentage referred to herein as X%. If the transactions are approved, the number of chargebacks that would have been received for the rejected transactions is not known, so the fraud rate for these transactions may have to be estimated and the percentage may simply be set to an estimated percentage. This estimate may be based on statistics for similar transactions that have been approved, or may be selected based on experience. For subsequent iterations, an initial fraud probability may be provided by the fraud probability module 54. [ If the decision to approve or deny the transaction is based on the results of the security check, the probability of fraud estimates can be taken into account.

Referring now to FIG. 6, a flow diagram illustrates a process 100 that may be executed by a transaction blocking system 50 to block transactions. At block 102, the process 100 may receive transaction data 58 for a new transaction. In response to receiving the transaction data 58, the process 100 may proceed to block 104 and estimate the fraud probability P _E for the transaction. If the process 100 still has to process any transactions in the current analysis period, the estimated fraud probability P _E can be set to the initial fraud probability P _I , as described above with respect to FIG. The initial fraud probability P _I may also be determined based on an initial or predetermined model that determines the fraud probability based on the transaction data 58. [ For example, the initial fraud probability P _I may be based on a combination of the average observed fraud rate for transactions approved during the previous accounting period, or the estimated fraud rate for transactions rejected during the previous accounting period. have. That is, in the fastest iteration of process 100, the data in transaction database 56 may be based on transactions in the case where a previous fraud blocking strategy is applied. In any event, in response to determining the estimated fraud probability P _E , the process 100 may proceed to block 106.

At block 106, the process 100 may determine the fraud cost C _F for the transaction based on the estimated fraud probability P _E. As described above with respect to FIG. 4, this determination may include generating a fraud cost C _F for each of a plurality of possible actions. Process 100 may then proceed to block 108 and make an initial determination as to which action to take by selecting an action that produces the lowest fraud cost C _F. In an embodiment of the invention, the available actions are: (1) approving the transaction; (2) requesting security checks and approving or rejecting transactions in response to the results; Or (3) rejecting the transaction without requesting security checks. For example, if the fraud cost to approve a transaction is less than the fraud cost of requesting a security check or refusing a transaction, then the initial judgment may be to approve the transaction. In response to the initial determination being determined, the process 100 may proceed to block 110. [

At block 110, the process 100 may determine whether the initial determination is to reject the transaction. In response to the initial determination not rejecting the transaction ("no" branch of decision block 110), process 100 may proceed to block 114. In response to the initial determination rejecting the transaction, the process 100 may proceed to block 112. At block 112, the process 100 may determine the expected impact on the cumulative fraud cost for a plurality of transactions by approving the transaction. If the impact on the cumulative fraud cost is greater than the fraud cost for the rejected transaction, the transaction blocking system 50 may be able to determine whether the current transaction . In some cases it may be useful to approve transactions that are normally rejected based on the fraud cost of the transaction, since data on the chargebacks are not collected on the transactions unless the transactions are approved. By approving a portion of these transactions, the transaction blocking system 50 can obtain chargeback data for trades that are typically rejected. This information can then be used to update fraud blocking parameters.

For example, for a particular transaction, the fraud cost C _F for approving the transaction is greater than the fraud cost C _F for rejecting the transaction, or

Where C _LS is the cost of lost sales resulting from an erroneously rejected transaction, C _P is the cost of processing the chargeback, and C _FL is the fraudulent liability cost (e.g., the cost of the travel service).

If the transaction is approved to acquire the information, the cost of approving the transaction to obtain this information, C _A , can be given by the difference between the fraud cost C _F authorizing the transaction and the fraud cost C _F denying the transaction: have:

It may be useful to approve a transaction if the expected cumulative fraud cost C _{F that} is saved in future transactions by obtaining information about the current transaction is greater than the approved cost C _A.

Judgments to allow the transaction in question can produce three scenarios regarding the configuration of the transaction training set: (1) the current training set (TR ₀ ) without additional transactions; (2) the current Training Set plus additional transaction (TR ₁ ) assuming that the additional transaction is deceptive; And (3) additional Training Set plus additional transaction (TR ₂ ) assuming that the additional transaction is not fraudulent. Determining the "gain" G in by approving the transaction due to the improvement in the accuracy of cumulative fraud cost: (1) the parameters for a fraud probability module 54 of each training set TR _0, TR _1, TR ₂ To decide; (2) calculating cumulative fraud costs C _F (TE, TR) for each set of parameters across the testing set TE; And (3) subtracting the cumulative sum of the cumulative fraud costs for the training sets, including additional transactions, from the cumulative fraud costs of the additional training-free training set. The gain G can be expressed in the form of the following equation:

Thus, the positive gain G may indicate an expected decrease in the cumulative fraud cost C _F (TE, TR) over the testing set TE resulting from adding the current transaction to the training set TR. If the gain G is greater than the transaction approval cost C _A for the transaction in question, the process 100 may approve the transaction before proceeding to block 114. If the gain G is not greater than the transaction approval cost C _A for the transaction in question, the process 100 may reject the transaction before proceeding to block 114.

At block 114, the process may update the estimated fraud probability P _E for transactions in the transaction database 56 based on additional information received after a previous update, such as information regarding chargebacks. Such updates may occur at regular intervals, such as once a day. The updated fraud probability P _E for rejected transactions may remain unchanged from the initial fraud probability P _I since new information about chargebacks may not be received for rejected transactions.

Each authorization transaction can have one of three states: (1) a chargeback has been received, in which case the probability of fraud P can be set to 100% for that transaction; (2) no chargeback has been received, in which case the estimated fraud probability P _E for that transaction may be reduced for each additional day passed without receiving a chargeback; And (3) the chargeback period has expired without receiving a chargeback, in which case the probability of fraud P may be set at 0% for that transaction. The decrease in the probability of fraud P for transactions that do not receive a chargeback but are still within the chargeback period can be estimated based on the data in the historical database as described above with respect to Figure 5 and is given by:

In certain tax jurisdictions, merchant fees may depend on the percentage of fraudulent transactions authorized by the merchant during the fiscal period. That is, the merchant fees may depend on the number of fraudulent transactions or the dollar amount and the ratio between the total sales during the accounting period in which the merchant fees are evaluated. To illustrate this change, the transaction blocking system 50 may consider the effects of fraudulent transactions on merchant fees. To this end, the transaction blocking system 50 may be configured to determine an anticipated merchant commission based on an estimate of sales by the merchant. This estimate may be provided by the merchant, or it may be estimated based on a set of any suitable parameters such as historical sales data for the merchant, time of year, and the like.

By way of example, by day n, which is the end of the accounting period, the transaction blocking system 50 may have approved m transactions during the accounting period. The transaction blocking system 50 can determine the cumulative expected chargeback amount CB _E for n days using the following equation:

Where P _i (n) is the updated fraud probability for transaction i for n days and C is the cost of transaction i, or the cost of the chargeback, one of which must be received.

For each partial-period (e.g., days) k of an accounting period (e.g., month, quarter, year), for each approved transaction, a batch run that updates the fraud probability is: (1) If the probability of fraud P is 1 (for example, a chargeback has been received), keep P equal to 1; (2) If a chargeback was received on day n, set P _i (n) = 1; (3) Otherwise, for each partial-period k,

To illustrate the current partial-period out-period, the transaction blocking system 50 performs a Monte-Carlo simulation to obtain the distribution of the expected charge rejection amounts over the accounting period . This simulation can be based, at least in part, on estimates of sales during the fiscal period. The anticipated merchant fee may then be determined based on the expected payoff volume distribution and estimated sales during the period.

In general, a sequence, or even a subset thereof, of routines, components, programs, objects, modules or instructions executed to implement embodiments of the invention, whether implemented as part of an operating system or a specific application, May be referred to herein as "computer program code" or simply "program code". The program code typically resides in various instances in various memory and storage devices in a computer and, when readable and executed by one or more processors in a computer, causes the computer to implement various aspects of embodiments of the invention And / or < / RTI > the elements necessary to execute the elements. The computer readable program instructions for carrying out the operations of the embodiments of the present invention may be source code or object code written in, for example, an assembly language or any combination of one or more programming languages.

The various program codes described herein may be identified based on the applications in which they are implemented in specific embodiments of the invention. However, any subsequent specific program nomenclature is used merely for convenience, and thus the present invention should not be limited to use in any particular application identified and / or implied by such nomenclature Should be recognized. Moreover, it is to be appreciated that the program functions may be implemented in a variety of software layers (e. G., ≪ RTI ID = 0.0 > For example, if the various schemes that can be assigned to an operating system, libraries, APIs, applications, applets, etc., are specified, then embodiments of the present invention may be used for specific organization and assignment of program functions But it should be recognized that the invention is not limited to these.

Program code embodied in any of the applications / modules described herein may be distributed individually or collectively as program objects in a variety of different forms. In particular, the program code may be distributed using a computer-readable storage medium having computer-readable program instructions thereon for causing the processor to perform aspects of embodiments of the present invention.

An essentially non-transitory computer-readable storage medium is any volatile and nonvolatile, implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules or other data, and Removable and non-separable tangible media. The computer readable storage medium may be any type of storage medium such as RAM, ROM, erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM) , Flash memory or other solid state memory technology, a portable compact disc read-only memory (CD-ROM), or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage, Or any other medium that can be used to store the desired information and that can be read by a computer. The computer-readable storage medium can be any type of computer-readable storage medium, such as electrical signals (e.g., electromagnetic waves propagating electromagnetic waves or other electromagnetic waves propagating through a transmission medium such as a waveguide) ) Should not be interpreted. The computer-readable program instructions may be downloaded from a computer, another type of programmable data processing apparatus, or from a computer-readable storage medium to another computer or external storage device via a network or other device.

Computer-readable program instructions stored on a computer-readable medium may be stored on a computer-readable medium such that instructions stored on the computer-readable medium may cause the computer to perform the functions, acts and / or actions specified in the flowcharts, sequence diagrams and / May be used to derive a computer, other types of programmable data processing apparatus, or other devices that function in a particular manner, to produce a manufacturing article containing the instructions to implement. Computer program instructions may include instructions for executing instructions through one or more processors to cause a series of computations to be performed on functions, operations and / or operations specified in flowcharts, sequence diagrams and / or block diagrams May be provided to a general purpose computer, special purpose computer, or other programmable data processing apparatus for producing a machine, so as to be implemented to implement the invention.

In certain alternative embodiments, the functions, acts, and / or operations specified in the flowcharts, sequence diagrams and / or block diagrams may be reordered, processed in series, and / or consistent with embodiments of the present invention And can be processed concurrently. Moreover, any of the flowcharts, sequence diagrams, and / or block diagrams may include more or fewer blocks than those consistent with the embodiments of the present invention.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the embodiments of the invention. As used herein, the singular forms "a," "an," and "the" are intended to also include the plural forms, unless the context clearly dictates otherwise. It will be further understood that the terms " comprises, " and / or "comprising" when used in this specification are taken to indicate the presence of stated features, integers, steps, operations, elements and / But do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. Moreover, the terms "includes", "having", "has", "with", "comprised of" To the extent that they are used in a range, such terms are intended to be inclusive in a manner similar to the term " comprising ".

While the invention has been illustrated by the description of various embodiments and these embodiments have been described in considerable detail, it is not the intention of the applicant to restrict or in any way limit the scope of the appended claims to such detail. Additional advantages and modifications will readily appear to those skilled in the art. The invention in its broader aspects is not therefore limited to the specific details, representative apparatus and method, and illustrative examples shown and described. Thus, deviations from such details can be made without departing from the spirit or scope of applicant's universal inventive concept.

Claims (39)

In a method for screening a transaction,
Receiving, at the computer, first data characterizing the first transaction;
Determining, by the computer, a first probability that the first transaction is fraudulent based on the first data;
Receiving, at the computer, second data characterizing a second transaction occurring during a period during which a chargeback for the first transaction can be received;
Determining, by the computer, a second probability of fraud of the first transaction based on the first probability and the amount of time since the approval of the first transaction; And
Determining, by the computer, a third probability of the second transaction being fraudulent based at least in part on the second data and the second probability;
The method comprising the steps of: The method according to claim 1,
Wherein determining the second probability comprises:
Determining a probability density function with respect to time to receive the chargeback for the first transaction; And
Determining an area of a portion of the probability density function corresponding to an amount of time since the approval of the first transaction;
Lt; / RTI >
And wherein the second probability is determined based at least in part on the area. 3. The method of claim 2,
Wherein the step of determining the probability density function comprises:
Determining a plurality of transactions in which chargebacks are received; And
Determining a reference distribution for the plurality of transactions with respect to an amount of time between the acceptance of each transaction and a receipt of a corresponding charge rejection
Lt; / RTI >
Wherein the probability density function is determined based on the reference distribution. 3. The method of claim 2,
Wherein the first transaction and the second transaction include a plurality of transactions by a merchant, each transaction having an accounting period comprising a plurality of sub-periods with a date of acceptance, Lt; / RTI >
The method comprises:
For each transaction in the period,
Determining whether the chargeback has been received;
Setting a fraud probability for the transaction to a second probability for the transaction if the chargeback has not been received;
Setting the fraud probability of the transaction to unity if the chargeback is received; And
Multiplying the fraud probability for the transaction by the cost of the transaction to produce an expected chargeback amount for the transaction
The method comprising the steps of: 5. The method of claim 4,
Wherein at least a portion of the plurality of transactions includes a transaction performed by an indirect seller,
The method comprises:
Determining a percentage of the transaction in the portion that is a card not present transaction; And
Adjusting the expected amount of chargeback for a transaction performed by the indirect seller based on the percentage
The method comprising the steps of: 5. The method of claim 4,
In response to receiving the chargeback, determining whether an underlying transaction is a card-non-existent transaction; And
Generating an agency debit memo to recover the chargeback in response to the base transaction being the card-less transaction;
The method comprising the steps of: 5. The method of claim 4,
Summing the expected chargeback amount to produce a cumulative expected chargeback amount; And
Determining a prospective merchant fee based on the cumulative expected chargeback amount
The method comprising the steps of: 8. The method of claim 7,
Further comprising determining a cost of fraud for the second transaction based on the expected chargeback amount for the second transaction and the expected merchant fee. 8. The method of claim 7,
Further comprising determining a chargeback amount density function for the merchant over the accounting period. 10. The method of claim 9,
Wherein the step of determining the chargeback amount density function comprises:
Summing the expected charge rejection amount of the plurality of transactions to generate a cumulative expected chargeback amount for each partial-period up to the current sub-period and including the current partial-period; And
Carlo simulation (Monte-Carlo simulation) to obtain the distribution of the expected charge rejection amount for each sub-period beyond the current sub-period
Lt; / RTI >
Wherein the charge reject quantity density function comprises the cumulative expected charge reject quantity. An apparatus for blocking transactions, comprising:
A processor; And
Memory containing instructions
Lt; / RTI >
Wherein the instructions, when executed by the processor, cause the device to:
To receive first data characterizing a first transaction,
Determine a first probability that the first transaction is fraudulent based on the first data,
To receive second data characterizing a second transaction occurring during a period during which a chargeback for the first transaction can be received,
Determine a second probability of the first transaction based on the first probability and the amount of time since the approval of the first transaction,
And cause the second transaction to determine a third probability of fraud based at least in part on the second data and the second probability. 12. The method of claim 11,
Wherein the instructions cause the device to:
Determining a probability density function with respect to time to receive the chargeback for the first transaction,
By determining an area of a portion of the probability density function corresponding to the amount of time since the approval of the first transaction,
And determine the second probability - the second probability is determined based at least in part on the area. 13. The method of claim 12,
Wherein the instructions cause the device to:
Determining a plurality of transactions for which chargebacks are received,
By determining a reference distribution for the plurality of transactions with respect to the amount of time between the acceptance of each transaction and the receipt of a corresponding charge rejection,
And determining a probability density function, the probability density function being determined based on the reference distribution. 13. The method of claim 12,
Wherein the first transaction and the second transaction include a plurality of transactions by a merchant, each transaction having an date of acceptance and an accounting period comprising a plurality of sub- Lt; / RTI >
The instructions also cause the device to, for each transaction in the accounting period,
To determine whether the chargeback has been received,
Set a fraud probability for the transaction to a second probability for the transaction if the chargeback has not been received,
If the payment rejection is received, the fraud probability for the transaction is set to 1 (unity)
And multiplies the fraud probability for the transaction by the cost of the transaction to produce an expected chargeback amount for the transaction. 15. The method of claim 14,
Wherein at least a portion of the plurality of transactions includes a transaction performed by an indirect seller, and wherein the instructions further cause the device to:
To determine the percentage of transactions in the portion of the card not present transaction
And adjust the expected amount of chargeback for a transaction performed by the indirect seller based on the percentage. 15. The method of claim 14,
The instructions may also cause the device to:
In response to receiving the charge rejection, to determine whether the ground transaction is a card non-existent transaction,
And in response to the base transaction being the card-less transaction, generating an agency debit memo to recover the chargeback. 15. The method of claim 14,
The instructions may also cause the device to:
Causing the expected chargeback amount to be added to produce a cumulative expected chargeback amount,
And determine an expected merchant fee based on the cumulative expected chargeback amount. 18. The method of claim 17,
The instructions may also cause the device to:
And determine a fraud cost for the second transaction based on the expected chargeback amount for the second transaction and the expected merchant fee. 18. The method of claim 17,
The instructions may also cause the device to:
And to determine a chargeback amount density function for the merchant over the accounting period. In a computer program product,
Non-transitory computer readable storage medium; And
Instructions stored on the non-transient computer-readable storage medium
Lt; / RTI >
Wherein the instructions, when executed by a processor, cause the processor to:
To receive first data characterizing a first transaction,
Determine a first probability of the first transaction based on the first data,
To receive second data characterizing a second transaction occurring during a period during which a chargeback for the first transaction can be received,
Determine a second probability of the first transaction based on the first probability and the amount of time since the approval of the first transaction,
Cause the second transaction to determine a third probability of fraud based at least in part on the second data and the second probability. In a transaction blocking method,
At the computer, receiving data defining a first transaction;
Determining, by the computer, a first fraud cost for taking a first action and a second fraud cost for taking a second action;
Determining a decrease in the cumulative fraud cost for a testing set of transactions that can be caused by taking said second action by said computer in response to said second fraud cost exceeding said first fraud cost ;
Determining, by the computer, a first difference between the first fraud cost and the second fraud cost; And
If the decrease is greater than the first difference, taking the second action by the computer
The method comprising the steps of: 22. The method of claim 21,
Wherein the first action includes rejecting the first transaction, and wherein the second action includes approving the first transaction. 22. The method of claim 21,
Wherein the step of determining a decrease in the cumulative fraud cost comprises:
Defining a training set of a plurality of transactions;
Determining a set of a plurality of parameters for a probability probability function, each set of parameters corresponding to and based on another set of the training sets;
Determining, for each set of parameters, the cumulative fraud cost for a testing set of the transaction; And
Subtracting a weighted sum of the first portion of the cumulative fraud cost from a second portion of the cumulative fraud cost to determine the decrease
The method comprising the steps of: 24. The method of claim 23,
Wherein the training set of the plurality of transactions comprises:
A first training set,
A second training set comprising the first training set and the first transaction that the first transaction assumes is fraud, and
The first training set including the first training set and the first transaction inferring that the first transaction is not a fraud,
The method comprising the steps of: 25. The method of claim 24,
Wherein determining the plurality of sets of parameters for the fraud probability function comprises:
Determining a first set of parameters for the fraud probability function based on the first training set;
Determining a second set of parameters for the fraud probability function based on the second set of training; And
Determining a third set of parameters for the fraud probability function based on the third training set
The method comprising the steps of: 26. The method of claim 25,
For each set of parameters, the step of determining the cumulative fraud cost for a testing set of the transaction comprises:
Determining a first cumulative fraud cost for the testing set using the first set of parameters;
Determining a second cumulative fraud cost for the testing set using the second set of parameters; And
Determining a third cumulative fraud cost for the testing set using the third set of parameters
The method comprising the steps of: 27. The method of claim 26,
Determining a probability that the first transaction is fraudulent; And
1 minus determining the second difference as the probability
Further comprising:
Wherein the weighted sum is equal to a product of the second cumulative fraud cost plus the probability plus a product of the third cumulative fraud cost and the second difference. 22. The method of claim 21,
Wherein the first transaction is performed by an indirect seller,
Determining a percentage of transactions by the indirect merchant that is a card not present transaction; And
Adjusting at least one of the first fraud cost or the second fraud cost for a transaction performed by the indirect seller based on the percentage
The method comprising the steps of: 22. The method of claim 21,
In response to receiving the charge rejection, determining whether the ground transaction is a card non-existent transaction; And
Generating an agency debit memo to recover the charge rejection in response to the base transaction being the card non-existent transaction;
The method comprising the steps of: An apparatus for blocking transactions, comprising:
A processor; And
Memory containing instructions
Lt; / RTI >
Wherein the instructions, when executed by the processor, cause the device to:
To receive data defining a first transaction,
To determine a first fraud cost for taking a first action and a second fraud cost for taking a second action,
Determine a reduction in the cumulative fraud cost for a testing set of transactions by taking the second action in response to the second fraud cost exceeding the first fraud cost,
Determine a first difference between the first fraud cost and the second fraud cost,
And if the decrease is greater than the first difference, take the second action. 31. The method of claim 30,
Wherein the first action includes rejecting the first transaction, and wherein the second action includes approving the first transaction. 31. The method of claim 30,
Wherein the instructions cause the device to:
Defining a training set of a plurality of transactions,
Determining a set of a plurality of parameters for a probability probability function, each set of parameters corresponding to and based on another set of the training sets;
Determining, for each set of parameters, the cumulative fraud cost for the testing set of the transaction;
By subtracting the weighted sum of the first portion of the cumulative fraud cost from a second portion of the cumulative fraud cost to determine the reduction,
And to determine a decrease in the cumulative fraud cost. 33. The method of claim 32,
Wherein the training set of the plurality of transactions comprises:
A first training set,
A second training set comprising the first training set and the first transaction that the first transaction assumes is fraud, and
A third training set including the first training set and the first transaction that the first transaction is not a fraud,
The device comprising: 34. The method of claim 33,
Wherein the instructions cause the device to:
Determining a first set of parameters for the fraud probability function based on the first training set,
Determining a second set of parameters for the fraud probability function based on the second training set,
And determining a third set of parameters for the fraud probability function based on the third training set,
And determine a set of the plurality of parameters for the fraud probability function. 35. The method of claim 34,
Wherein the instructions cause the device to:
Determining a first cumulative fraud cost for the testing set using the first set of parameters,
Determining a second cumulative fraud cost for the testing set using the second set of parameters,
By determining a third cumulative fraud cost for the testing set using the third set of parameters,
And for each set of parameters to determine the cumulative fraud cost for the testing set of the transaction. 36. The method of claim 35,
The instructions may also cause the device to:
Determine a probability that the first transaction is fraudulent,
1 minus to determine the second difference as the probability,
Wherein the weighted sum is equal to a product of the second cumulative fraud cost plus the probability plus a product of the third cumulative fraud cost and the second difference. 31. The method of claim 30,
Wherein the first transaction is performed by an indirect seller,
The instructions may also cause the device to:
To determine the percentage of transactions by the indirect seller that is a card not present transaction,
And to adjust at least one of the first fraud cost or the second fraud cost for a transaction performed by the indirect seller based on the percentage. 31. The method of claim 30,
Wherein the instructions cause the device to:
In response to receiving the charge rejection, determine whether the ground transaction is a card non-existent transaction,
And to generate an agency debit memo to recover the chargeback in response to the base transaction being the card non-existent transaction. In a computer program product,
Non-transitory computer readable storage medium; And
Instructions stored on the non-transient computer-readable storage medium
Lt; / RTI >
Wherein the instructions, when executed by a processor, cause the processor to:
To receive data defining a first transaction,
To determine a first fraud cost for taking a first action and a second fraud cost for taking a second action,
Determine a reduction in the cumulative fraud cost for a testing set of transactions by taking the second action in response to the second fraud cost exceeding the first fraud cost,
Determine a first difference between the first fraud cost and the second fraud cost,
And if the decrease is greater than the first difference, take the second action.

Images