Electronic transactional data based predictive system

• US 10,387,883 B2
• Filed: 02/09/2017
• Issued: 08/20/2019
• Est. Priority Date: 02/09/2017
• Status: Active Grant

• Current Assignee: Visa International Service Association (Visa, Inc.)
• Original Assignee: Visa International Service Association (Visa, Inc.)
• Inventors: Wang, Guangyu, Lu, Kebin, Hwang, Sang, Kumar, Debesh, Jindal, Preeti, Lau, Lawson, Tran, Melissa Lawu, Giglio, Japneet Kaur
• Primary Examiner(s): Zeender, Florian M
• Assistant Examiner(s): Deligi, Vanessa
• Application Number:
  US15/428,909
• Publication Number:
  US 20180225667A1
• Time in Patent Office: 922 Days
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/20 Ensemble learning
• CPC Class Codes: G06N 5/01 Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: G06Q 20/4016 involving fraud or risk level assessment in transaction processing
• CPC Class Codes: G06Q 40/12 Accounting
• CPC Class Codes: H04W 8/00 Network data management

Abstract

Past electronic records may be studied using a computer learning algorithm in order to make predictions of future use of the electronic accounts.

Claims

1. A computer system for determining usage change predictions for electronic accounts comprising:

a transaction server physically configured to;

obtain purchase data wherein purchase data comprises past data of purchases by a plurality of electronic accounts;

a learning server physically configured to;

execute an attrition learning algorithm to determine attrition elements that predict likelihood of attrition in a near future;

execute a transaction reduction algorithm to determine transaction reduction elements that predict likelihood of transaction reductions in the near future;

a reviewing server physically configured to;

select accounts of the plurality of electronic accounts that contain the attrition elements determined to predict likelihood of attrition in the near future;

assign an attrition score to the accounts selected for containing the attrition elements determined to predict likelihood of attrition in the near future;

select accounts of the plurality of electronic accounts that contain the transaction reduction elements;

assign a transaction reduction score to the accounts selected for containing the transaction reduction elements;

an analysis server physically configured to;

provide a user interface for selecting a transaction value threshold, an attrition score threshold, and a transaction reduction score threshold, the user interface being configured to provide a graphical representation of a number of accounts meeting each threshold;

determine high value accounts wherein the high value accounts have a value of transactions over the transaction value threshold;

select high risk accounts from the high value accounts wherein the high risk accounts have attrition scores above the attrition score threshold and transaction reduction scores above the transaction reduction score threshold;

in response to an account being selected as being high risk, select a message to be communicated to the high risk accounts; and

communicate the selected message to the high risk accounts.

2. The computer system of claim 1, wherein the analysis server is further configured to:

separate the accounts by spend amount into a plurality of spend levels based on spend;

separate the accounts by transaction reduction score into a plurality of reduction levels;

separate the accounts by attrition score into a plurality of attrition levels; and

for each spend level, create a matrix with plurality attrition on a first axis, plurality of reductions levels on a second axis and spend levels on a third axis.

3. The computer system of claim 1, wherein the message further comprises an enticement to use the electronic account.

4. The computer system of claim 3, wherein the enticement comprises at least one selected from a group comprising:

a higher credit limit, bonus rebates, higher percentage cash back, lower interest rates.

5. The computer system of claim 1, wherein the learning algorithm in the learning server determines an optimal time to communicate the enticement.

6. The computer system of claim 1, wherein the attrition learning algorithm reviews a plurality of elements during a study time period of n month intervals wherein the elements comprise at least one element selected from a group comprising:

total present value in month n of the study period;

percent of change in transactions for month n of study period over a last month of previous time period;

percent of change in transaction for month n over the first month during study period;

month-over-month change in everyday present value for month n over month 1 during the study period;

total online present value for a second quarter of the study period;

percent of change in recurring present value for month n-2 of study period over month n-2 of previous study period;

total number of decline during the study window;

month-over-month change in present value month n in the study period over month n of prior n months period;

total monthly charge count during the study period;

month-over-month change in transaction month n in the study period over month n of the prior n months period;

percent of change in transaction first month of study period over month n of previous n months; and

month-over-month change in recurring transaction for month n in the study period over month n of the prior n months period.

7. The computer system of claim 1, wherein action is taken within two time periods of the study period.

8. The computer system of claim 1, wherein the learning server uses a learning algorithm to determine transaction reduction based on:

change in everyday transaction on in second quarter vs first quarter during the study period;

percent change in transaction 2 quarter vs 1st quarter during the study period;

month over month change in transaction for month 6 of the study period;

month over month change in everyday transaction for month 6 of the study period;

month over month change in everyday transaction for month 5 of the study period;

number of decline during the study period; and

month over month change in online transaction for month 5 during the study period.

9. The computer system of claim 1, wherein the purchase data is card specific.

10. The computer system of claim 1, wherein the purchase data is merchant specific.

11. The computer system of claim 1, wherein the attrition learning algorithm uses a Treenet based learning algorithm.

12. A computer based method for determining usage change predictions for electronic accounts comprising:

obtaining purchase data from a transaction server wherein purchase data comprises past data of purchases by a plurality of electronic accounts;

executing an attrition learning algorithm in a learning server to determine attrition elements that predict likelihood of attrition in a near future;

executing a transaction reduction algorithm in the learning server to determine transaction reduction elements that predict likelihood of transaction reductions in the near future;

selecting accounts of the plurality of electronic accounts in a reviewing server that contain the attrition elements determined to predict likelihood of attrition in the near future;

assigning an attrition score to the accounts in the reviewing server selected for containing the attrition elements determined to predict likelihood of attrition in the near future;

selecting accounts of the plurality of electronic accounts in the reviewing server that contain the transaction reduction elements;

assigning a transaction reduction score to the accounts in the reviewing server selected for containing the transaction reduction elements;

providing a user interface in an analysis server for selecting a transaction value threshold, an attrition score threshold, and a transaction reduction score threshold, the user interface being configured to provide a three-dimensional graphical representation of a number of accounts meeting the selected thresholds;

determining high value accounts in the analysis server wherein the high value accounts that have a value of transactions over the transaction value threshold;

selecting high risk accounts in the analysis server from the high value accounts wherein the high risk accounts have attrition scores above the attrition score threshold and transaction reduction scores above the transaction reduction score threshold;

in response to an account being selected as being high risk, selecting in the analysis server a message to be communicated to the high risk accounts; and

communicating from the analysis server the selected message to the high risk accounts.

13. The method of claim 12, further comprising:

separate the accounts by spend amount into a plurality of spend levels based on spend;

separate the accounts by transaction reduction score into a plurality of reduction levels;

separate the accounts by attrition score into a plurality of attrition levels; and

for each spend level, create a matrix with plurality attrition on a first axis, plurality of reductions levels on a second axis and spend levels on a third axis.

14. The method of claim 12, wherein the message further comprises an enticement to use the electronic account.

15. The method of claim 14, wherein the learning algorithm in the learning server determines an optimal time to communicate the enticement.

16. The method of claim 12, wherein the learning attrition algorithm reviews a plurality of elements during a study time period of n month intervals wherein the elements comprise at least one element selected from a group comprising:

total present value in month n of the study period;

percent of change in transactions for month n of study period over the last month of previous time period;

percent of change in transaction for month n over a first month during study period;

month-over-month change in everyday present value for month n over month 1 during the study period;

total online present value for the second quarter of the study period;

percent of change in recurring present value for month n-2 of study period over month n-2 of previous study period;

total number of decline during the study window;

month-over-month change in present value month n in the study period over month n of the prior n months period;

total monthly charge count during the study period;

month-over-month change in transaction month n in the study period over month n of the prior n months period;

percent of change in transaction first month of study period over the month n of previous n months; and

month-over-month change in recurring transaction for month n in the study period over month n of the prior n months period.

17. The method of claim 16, wherein the learning server uses a learning algorithm to determine transaction reduction based on:

change in everyday transaction in second quarter vs first quarter during the study period;

percent change in transaction second quarter vs first quarter during the study period;

month over month change in transaction for month 6 of the study period;

month over month change in everyday transaction for month 6 of the study period;

month over month change in everyday transaction for month 5 of the study period;

number of decline during the study period; and

month over month change in online transaction for month 5 during the study period.

18. The method of claim 12, wherein the purchase data is card specific or merchant specific.

19. The method of claim 12, wherein the attrition learning algorithm uses a Treenet based learning algorithm.