US-12619894-B2 - Processing machine learning attributes

US12619894B2US 12619894 B2US12619894 B2US 12619894B2US-12619894-B2

Abstract

Systems and methods for processing machine learning attributes are disclosed. An example method includes: identifying a user transaction associated with a set of transaction attributes and a first transaction status; selecting, based on a risk evaluation model, a first plurality of transaction attributes from the set of transaction attributes; modifying a first value of a first transaction attribute in the first plurality of transaction attributes to produce a first modified plurality of transaction attributes; determining, based on the risk evaluation model, that the first modified plurality of transaction attributes identify a second transaction status different from the first transaction status; and in response to the determining, identifying the first transaction attribute as a risk attribute associated with the user transaction.

Inventors

Ying Lin
Huagang Yin
Jiaqi Zhang

Assignees

PAYPAL, INC.

Dates

Publication Date: 20260505
Application Date: 20230607

Claims (20)

1 . A computer system, comprising: a hardware processor; and a non-transitory computer-readable medium having stored thereon instructions that are executable to cause the computer system to perform operations comprising: receiving information indicating a first request to perform an electronic file transfer transaction associated with transferring electronic data between a first computing device and a second computing device via a network, wherein the information comprises a plurality of attribute values corresponding to a plurality of transaction attributes and associated with the electronic file transfer transaction; providing the plurality of attribute values as input values to a machine learning model, wherein the machine learning model generates an output usable to process the electronic file transfer transaction based on the input values; determining that the electronic file transfer transaction is denied or delayed based on the output of the machine learning model exceeding a specified amount; subsequent to the determining that the electronic file transfer transaction is denied or delayed, establishing a chat session with the first computing device; obtaining an utterance provided by a user of the first computing device during the chat session; selecting, from the plurality of transaction attributes, a subset of transaction attributes to be evaluated based on analyzing training data used to train the machine learning model and analyzing words within the utterance; determining, from a subset of the plurality of attribute values corresponding to the subset of transaction attributes, a particular attribute value corresponding to a network address of the first computing device that contributes to the output of the machine learning model exceeding the specified amount based on iteratively (i) modifying different attribute values from the subset of the plurality of attribute values and (ii) providing the modified attribute values to the machine learning model; in response to determining that the particular attribute value corresponding to the network address of the first computing device contributes to the output exceeding the specified amount, providing, to the first computing device, an indication of the particular attribute value that enables the first computing device to modify the network address associated with the first computing device; and in response to receiving, from the first computing device, a second request comprising the modified network address associated with the first computing device, authorizing the electronic file transfer transaction based on the modified network address.
2 . The computer system of claim 1 , wherein the operations further comprise: in response to determining the particular attribute value corresponding to the network address of the first computing device that contributes to the output of the machine learning model exceeding the specified amount, transmitting, to the first computing device, instructions for modifying the network address of the first computing device.
3 . The computer system of claim 1 , wherein the output is a first output of the machine learning model, wherein the machine learning model is configured to generate a second output based on the modified network address, and wherein the authorizing the electronic file transfer transaction is further based on the second output not exceeding the specified amount.
4 . The computer system of claim 1 , wherein the operations further comprise: generating an indication that explains one or more reasons for the electronic file transfer transaction being denied or delayed based on the determining that the particular attribute value contributed to the output exceeding the specified amount; and transmitting the indication to at least one of the first computing device or the second computing device.
5 . The computer system of claim 1 , wherein the iteratively modifying comprises modifying different ones of the different attribute values in each iteration of a plurality of iterations.
6 . The computer system of claim 1 , wherein the plurality of transaction attributes comprises a user transaction frequency level.
7 . A method, comprising: receiving, at a computer system, a first request to perform a data transfer between a first computing device and a second computing device, wherein the first request is associated with a plurality of attribute values corresponding to a plurality of attributes related to the data transfer; determining, using the plurality of attribute values as input values for a machine learning model, that the data transfer is denied or delayed based on an output of the machine learning model exceeding an amount, wherein the machine learning model generates the output based on the input values; selecting, from the plurality of attributes, a subset of attributes to be evaluated based on analyzing training data used to train the machine learning model; determining, from a subset of the plurality of attribute values corresponding to the subset of attributes, a particular attribute value corresponding to a network address of the first computing device that contributes to the output of the machine learning model exceeding the amount based on (i) modifying different attribute values from the subset of the plurality of attribute values and (ii) providing the modified attribute values to the machine learning model; providing, to the first computing device, an indication that enables the first computer device to modify the network address of the first computing device; receiving a second request comprising the modified network address associated with the first computing device; and processing the second request based on the modified network address.
8 . The method of claim 7 , wherein the plurality of transaction attributes includes a transaction frequency level attribute associated with one of the first computing device or the second computing device.
9 . The method of claim 7 , wherein the plurality of attributes includes a device location attribute associated with one of the first computing device or the second computing device.
10 . The method of claim 7 , wherein the plurality of attributes includes a description attribute of data associated with the data transfer.
11 . The method of claim 7 , further comprising: generating information that provides one or more reasons for the data transfer being denied or delayed based on determining that the particular attribute value contributed to the output exceeding the amount; and transmitting the information to the first computing device, wherein the first computing device corresponds to a requestor of the first request.
12 . The method of claim 7 , further comprising: transmitting an indication of the data transfer being denied or delayed to at least one of the first computing device or the second computing device.
13 . The method of claim 11 , wherein the generating the information is further based on at least a second attribute value of the plurality of attribute values.
14 . A non-transitory machine-readable medium having instructions stored thereon that, in response to being executed by one or more processors, cause a system to perform operations comprising: receiving a first request to perform a data transfer between a first computing device and a second computing device, wherein the first request is associated with a plurality of attribute values corresponding to a plurality of attributes and related to the data transfer; providing the plurality of attribute values as input values to a machine learning model, wherein the machine learning model is configured to generate an output usable to process the first request based on the input values; determining that the data transfer is denied or delayed based on the output of the machine learning model exceeding a threshold; selecting, from the plurality of attributes, a subset of attributes to be evaluated based on analyzing training data used to train the machine learning model; determining, from the plurality of attribute values provided to the machine learning model, a particular attribute value corresponding to a computer configuration of the first computing device that contributes to the output exceeding the threshold based on (i) modifying different attribute values from the subset of the plurality of attribute values and (ii) providing the modified attribute values to the machine learning model; providing, to the first computing device, an indication that the particular attribute value contributes to a denial or a delay of the data transfer, wherein the indication enables the first computing device to modify the computer configuration; and in response to receiving, from the first computing device, a second request comprising the modified computer configuration of the first computing device, authorizing the data transfer based on the modified computer configuration of the first computing device.
15 . The non-transitory machine-readable medium of claim 14 , wherein the operations further comprise: in response to determining the particular attribute value corresponding to the computer configuration of the first computing device that contributes to the output exceeding the threshold, transmitting, to the first computing device, instructions for modifying the computer configuration.
16 . The non-transitory machine-readable medium of claim 14 , wherein the operations further comprise: generating content comprising one or more reasons for the data transfer being denied or delayed based on the determining that the particular attribute value contributed to the output exceeding the threshold; and transmitting the content to the first computing device, wherein the first computing device corresponds to a requestor of the first request.
17 . The non-transitory machine-readable medium of claim 14 , wherein the operations further comprise: transmitting an indication of the data transfer being denied or delayed to at least one of the first computing device or the second computing device.
18 . The non-transitory machine-readable medium of claim 14 , wherein the request is received via the second computing device, and wherein the second computing device corresponds to a recipient of the data transfer.
19 . The non-transitory machine-readable medium of claim 14 , wherein the plurality of transaction attributes includes an IP address attribute associated with one of the first computing device or the second computing device.
20 . The non-transitory machine-readable medium of claim 14 , wherein the plurality of transaction attributes includes a device location attribute associated with one of the first computing device or the second computing device.

Description

CROSS REFERENCE TO RELATED APPLICATIONS The present application is a continuation application of U.S. patent application Ser. No. 17/681,480, filed Feb. 25, 2022, which is a continuation of U.S. patent application Ser. No. 15/296,520, filed Oct. 18, 2016, now U.S. Pat. No. 11,301,765, and they are now incorporated in reference in their entirety. TECHNICAL FIELD The present disclosure relates generally to machine learning, and in particular, to processing machine learning attributes. BACKGROUND Entities transacting with users often delay or deny high risk transactions to avoid potential loss. For example, an attempted transfer of computer files to an unregistered user device may suggest that the file transfer is unauthorized, and an attempted user authentication on a desktop computer having a never-seen-before IP address may indicate that the user authentication is fraudulent. When delaying or denying such a transaction, a merchant often does not provide any specific explanation to a user attempting the transaction. For example, an online merchant may not inform a user why her purchase request made using a friend's computer and the friend's payment account was denied unless the user calls into the merchant's customer support center to inquire. Because transaction delays or denials are determined based on risk calculations using one or more risk evaluation models, investigating on which basis a transaction was denied or delayed may therefore require querying a risk determination model. A risk determination model, however, may employ hundreds or even thousands of variables. It can therefore be time- and resource-consuming (e.g., may take days and several computing servers) to identify specific reasons for a transaction denial. Also, a merchant's customer support team may have to relay customer inquiries to and rely on a risk assessment team, as the support team may not possess the technical expertise to query the risk evaluation model themselves. Customer experience may be enhanced when the support staff can articulate a reason in a timely fashion. There is therefore a need for a device, system, and method, which processes user queries of machine learning attributes and returns query results with reasonable certainty and within a predefined time frame. BRIEF DESCRIPTION OF THE FIGURES FIG. 1 is a schematic view illustrating an embodiment of a system for selecting machine learning attributes and modifying the values thereof to determine whether an attribute contributed to a transaction denial or delay. FIG. 2A-2B is a flow chart illustrating an embodiment of a process for selecting machine learning attributes and modifying the values thereof to determine whether an attribute contributed to a transaction denial or delay. FIG. 3 is a flow chart illustrating an embodiment of a method for selecting machine learning attributes and modifying the values thereof to determine whether an attribute contributed to a transaction denial or delay. FIG. 4 is a schematic view illustrating an embodiment of a user device. FIG. 5 is a schematic view illustrating an embodiment of a computing system. Embodiments of the present disclosure and their advantages are best understood by referring to the detailed description that follows. It should be appreciated that like reference numerals are used to identify like elements illustrated in one or more of the figures, wherein showings therein are for purposes of illustrating embodiments of the present disclosure and not for purposes of limiting the same. DETAILED DESCRIPTION The present disclosure provides systems and methods for selecting machine learning attributes and modifying the values thereof to detect whether a particular attribute may have contributed to a transaction denial or delay. To determine the basis on which a user transaction is denied (or delayed), in some implementations, a computing system may first compare transaction attributes (e.g., the total payment amount, the payment method used, and the IP address of the user device on which the transaction was requested) with training data attributes (the transaction attributes that were used to construct a risk evaluation model against which the user is querying) to identify matching transaction data attributes. This comparison process may narrow down the identification of relevant attributes from hundreds or thousands of attributes used in the risk assessment model to a few hundred matching attributes. The computing system may then modify the values of the selected transaction attributes to produce modified transaction data, for example, changing the network location of the user device used to request the user transaction from an unverified VPN address (e.g., 225.225.0.1) to a verified non-VPN address (e.g., 168.42.122.964) or changing the total transaction amount (e.g., from $100 to $20). The computing system next feeds the modified transaction data to the risk evaluation model to evaluate the risk level of the modified tra