Systems and methods for failed payment recovery systems
Systems and methods for failed payment recovery can include a method for payment recovery. The method includes steps for receiving a set of transaction information, predicting a set of one or more authorization field values based on the received set of transaction information, predicting a set of one or more optimal send times based on the received set of transaction information, and transmitting a set of one or more authorization messages based on the predicted set of authorization field values and the predicted set of optimal send times.
1. A method for payment recovery, the method comprising:
receiving, at a computing device, a set of transaction information;
predicting, using a set of processors in the computing device, a set of one or more authorization field values based on the received set of transaction information, wherein the set of one or more authorization field values is predicted by using a set of one or more machine learning models to determine a subset of the set of one or more authorization field values to be included in an authorization message to produce a highest likelihood of successful authorization,
wherein predicting the set of one or more authorization field values to be included in an authorization message to produce a highest likelihood of successful authorization comprises filtering out one or more authorization field values that decrease the likelihood of successful authorization,
wherein predicting the set of one or more authorization field values comprises modifying at least one authorization field value of the set of one or more authorization field values;
predicting, using the set of processors in the computing device, a set of one or more optimal send times based on the received set of transaction information, wherein the set of one or more optimal send times is predicted by using the set of one or more machine learning models to determine a set of one or more optimal send times at which to send an authorization message to produce a highest likelihood of successful authorization,
wherein the set of one or more optimal send times is predicted based on an expected number of transactions at a given credit card issuer for an authorization message,
wherein predicting the set of one or more optimal send times comprises using a reinforcement learning approach; and
transmitting, from the computing device to the given credit card issuer, a set of one or more authorization messages based on the predicted set of one or more authorization field values and the predicted set of one or more optimal send times.
2. The method of claim 1 , wherein the transaction information comprises transaction amount, card information, merchant information, and product information.
3. The method of claim 1 , wherein predicting the set of one or more authorization field values comprises filtering to exclude at least one authorization field value from the set of one or more authorization field values.
4. The method of claim 1 , wherein the set of authorization messages are formatted in accordance with ISO-8583.
5. The method of claim 1 further comprising:
determining that an authorization message has failed;
predicting a second set of one or more authorization field values; and
transmitting a second set of one or more authorization messages based on the predicted second set of one or more authorization field values.
6. The method of claim 1 further comprising:
determining that an authorization message has failed;
predicting a second set of one or more optimal send times; and
transmitting a second set of one or more authorization messages based on the predicted second set of one or more optimal send times.
7. The method of claim 1 , wherein predicting the set of one or more authorization field values to be included in an authorization message to produce a highest likelihood of successful authorization further comprises varying transaction amounts to be authorized based on previous authorization attempts.
8. The method of claim 1 , wherein predicting the set of one or more optimal send times further comprises generating a dunning schedule for a particular customer based on the predicted set of optimal send times.
9. The method of claim 1 , wherein the reinforcement learning approach comprises an epsilon greedy explore exploit approach.
10. A non-transitory computer readable medium containing processor instructions for payment recovery, where execution of the instructions by a processor causes the processor to perform a process that comprises:
receiving, at a computing device, a set of transaction information;
predicting, using a set of processors in the computing device, a set of one or more authorization field values based on the received set of transaction information, wherein the set of one or more authorization field values is predicted by using a set of one or more machine learning models to determine a subset of the set of one or more authorization field values to be included in an authorization message to produce a highest likelihood of successful authorization,
wherein predicting the set of one or more authorization field values to be included in an authorization message to produce a highest likelihood of successful authorization comprises filtering out one or more authorization field values that decrease the likelihood of successful authorization,
wherein predicting the set of one or more authorization field values comprises modifying at least one authorization field value of the set of one or more authorization field values;
predicting, using the set of processors in the computing device, a set of one or more optimal send times based on the received set of transaction information, wherein the set of one or more optimal send times is predicted by using the set of one or more machine learning models to determine a set of one or more optimal send times at which to send an authorization message to produce a highest likelihood of successful authorization,
wherein the set of one or more optimal send times is predicted based on an expected number of transactions received by a given credit card issuer for an authorization message,
wherein predicting the set of one or more optimal send times comprises using a reinforcement learning approach; and
transmitting, from the computing device to the given credit card issuer, a set of one or more authorization messages based on the predicted set of one or more authorization field values and the predicted set of one or more optimal send times.
11. The non-transitory computer readable medium of claim 10 , wherein the transaction information comprises transaction amount, card information, merchant information, and product information.
12. The non-transitory computer readable medium of claim 10 , wherein predicting the set of one or more authorization field values comprises at least one of filtering to exclude at least one authorization field value from the set of one or more authorization field values.
13. The non-transitory computer readable medium of claim 10 , wherein the process further comprises:
determining that an authorization message has failed;
predicting a second set of one or more authorization field values; and
transmitting a second set of one or more authorization messages based on the predicted second set of one or more authorization field values.
14. The non-transitory computer readable medium of claim 10 , wherein the process further comprises:
determining that an authorization message has failed;
predicting a second set of one or more optimal send times; and
transmitting a second set of one or more authorization messages based on the predicted second set of one or more optimal send times.
15. The non-transitory computer readable medium of claim 10 , wherein predicting the set of one or more authorization field values to be included in an authorization message to produce a highest likelihood of successful authorization further comprises varying transaction amounts to be authorized based on previous authorization attempts.
16. The non-transitory computer readable medium of claim 10 , wherein predicting the set of one or more optimal send times further comprises generating a dunning schedule for a particular customer based on the predicted set of optimal send times.
17. The non-transitory computer readable medium of claim 10 , wherein the reinforcement learning approach comprises an epsilon greedy explore exploit approach.