| CPC G06Q 20/401 (2013.01) [G06Q 20/085 (2013.01); G06Q 20/40 (2013.01)] | 19 Claims |
|
1. A computing platform comprising:
a network interface;
at least one processor;
at least one non-transitory computer-readable medium; and
program instructions stored on the at least one non-transitory computer-readable medium that are executable by the at least one processor such that the computing platform is configured to:
identify a pool of historical transactions involving multiple different customer accounts and multiple different merchants;
generate a training dataset comprising, for each historical transaction in the pool of historical transactions, (i) feature data for the historical transaction and (ii) an indicator of whether or not the historical transaction is a recurring charge;
apply a supervised machine learning process to the training dataset and thereby train a machine-learning model that functions to predict whether a transaction involving a customer account and a merchant is a recurring charge without relying on any encoded recurring transaction indicator for the transaction by (a) evaluating feature data for the transaction, wherein the evaluated feature data comprises values for (i) at least one transaction-level feature that is derived based on transaction-level data for the transaction and is independent from any recurring transaction indicator that is encoded within the transaction-level data for the transaction, (ii) at least one merchant-level feature that is derived based on historical transaction data for past transactions involving the merchant, and (iii) multiple account-merchant-level features that are derived based on historical transaction data for past transactions involving both the customer account and the merchant, and (b) based on the evaluation of the feature data, outputting a score for the transaction that indicates a predicted likelihood that the transaction is a recurring charge, wherein:
the at least one transaction-level feature comprises at least one of (1) an indication of whether a transaction time of the transaction was within a defined window of time or (2) an indication of whether a transaction amount of the transaction was within a defined range of amounts;
the at least one merchant-level feature comprises at least one of (1) an indication of a size of the merchant involved in the transaction, (2) an indication of a concentration of the merchant's historical transactions with respect to transaction dates, transaction times, or transaction amounts, or (3) an indication of an extent of the merchant's historical transactions that share at least one common characteristic with the transaction; and
the multiple account-merchant-level features comprise at least (1) an indication of an extent of historical transactions involving both the customer account and the merchant that follow a first type of frequency-based transaction pattern and (2) an indication of an extent of historical transactions involving both the customer account and the merchant that follow a second type of frequency-based pattern, wherein the second type of frequency-based pattern differs from the first type of frequency-based transaction pattern;
for each respective transaction of a plurality of transactions involving multiple different customer accounts and multiple different merchants:
obtain data for a respective transaction involving a respective customer account and a respective merchant that has been previously processed within a networked payment environment;
apply pre-processing logic to the obtained data for the respective transaction and thereby derive feature data for the respective transaction that includes respective values for (i) the at least one transaction-level feature, (ii) the at least one merchant-level feature, and (iii) the multiple account-merchant-level features;
input the feature data for the respective transaction into the trained machine-learning model and thereby determine a score for the respective transaction that indicates a likelihood that the respective transaction is a recurring charge; and
based on the score for the respective transaction, determine whether to classify the respective transaction as a recurring charge;
after classifying a given transaction involving a given customer account and a given merchant as a recurring charge, cause an account holder of the given customer account to be presented with a notification that the given transaction has been classified as a recurring charge, wherein the notification comprises one of a mobile push notification, an email notification, or a text-message notification, wherein the notification is presented via a networked end-user device associated with the account holder, and wherein the notification includes a selectable element for accessing a recurring-charges dashboard;
receive, from the networked end-user device associated with the account holder, an indication that the account holder has selected the selectable element for accessing the recurring-charges dashboard that was included in the notification; and
in response to receiving the indication, cause the account holder of the given customer account to be presented with an instance of the recurring-charges dashboard that comprises a listing of transactions involving the given customer account that have been classified as recurring charges, wherein each respective transaction in the listing of transactions is associated with a selectable element that, when selected, causes the instance of the recurring-charges dashboard to present any one or more other transactions involving the given customer account and a same merchant as the listed transaction that are determined to be related to the respective transaction.
|