Neural network learning for the prevention of false positive authorizations
Disclosed herein are systems and methods that identify and improve upon false positive scores. Some embodiments may include artificial neural network learning methods that utilize data input from users as well as enterprise machines. Information, such as transaction timing, prior transaction data, and demographics, may be taken as inputs to the neural network processing. The processing may be done on one or more neural network computers or nodes.
1. A method of machine learning in an Artificial Neural Network (ANN) for prevention of false positives, the method comprising:
upon receiving an indication that a transaction has been declined based on one or more rules, receiving, by an ANN server, demographic information from a user terminal about a user associated with a user account in the transaction;
retrieving, by the ANN server, information from an enterprise terminal regarding a user profile, wherein the information represents transaction history of the user;
updating, by the ANN server, values on each neural network node of the ANN based on the demographic information from the user terminal and the transaction history information from the enterprise terminal;
upon updating the values on each neural network node, generating, by the ANN server, a false positive score based on a weighting of the values on each neural network node; and
transmitting, by the ANN server, the false positive score to the enterprise terminal, wherein the enterprise terminal is configured to display a graphical user interface populated with the false positive score, and wherein upon receiving the false positive score, the enterprise terminal generates a new version of the one or more rules based on the false positive score and executes the new version of the one or more rules for a new transaction.
2. The method of claim 1 , wherein calculating the false positive score accounts for a time difference between the declined transaction and subsequent transactions.
3. The method of claim 1 , wherein generating the false positive score accounts for a difference in cost between the declined transaction and subsequent transactions.
4. The method of claim 1 , wherein generating the false positive score accounts for whether a channel of the declined transaction and subsequent transactions is the same.
5. The method of claim 1 , wherein generating the false positive score accounts for demographics information of the user for a card making a subsequent transaction and the user of the declined transaction including age, income, and gender.
6. The method of claim 1 , wherein generating the false positive score accounts for the user profile of the user for a card making a subsequent transaction and the profile of the user of the declined transaction.
7. The method of claim 1 , wherein generating the false positive score includes utilizing the following formula:
False Positive Score (FPS)=α+β1*Same/Similar Merchant+β2*Time difference+/β3*Amount difference+/β4*Same/Different channel+β5*similar demographics+β6*similar profile,
where α=fixed component determined from False positive data of issuer, values of coefficients β1, β2, β3, β4, β5, and β 6 are based on assigned significances of predicting variables to predict a probability of a false positive, same/Similar Merchant=0 or 1, same/Different channel=0 or 1, similar demographics=0 or 1, and similar profile=a score between 0 and 1 depending on similarity in profile of a declined card and a card making a next transaction.
8. A server configured for machine learning in an Artificial Neural Network (ANN) for prevention of false positives, the server comprising:
a memory; and
a processor configured to:
upon receiving an indication that a transaction has been declined based on one or more rules, receive demographic information from a user terminal about a user associated with a user account in the transaction;
retrieve information from an enterprise terminal regarding a user profile, wherein the information represents transaction history of the user;
update values on each neural network node of the ANN based on the demographic information from the user terminal and the transaction history information from the enterprise terminal;
upon updating the values on the neural network node, generate a false positive score based on a weighting of the values on each neural network node; and
transmit the false positive score to the enterprise terminal, wherein the enterprise terminal is configured to display a graphical user interface populated with the false positive score, wherein upon receiving the false positive score, the enterprise terminal generates a new version of the one or more rules based on the false positive score and executes the new version of the one or more rules for a new transaction.
9. The server of claim 8 , wherein generating the false positive score accounts for a time difference between the declined transaction and subsequent transactions.
10. The server of claim 8 , wherein generating the false positive score accounts for a difference in cost between the declined transaction and subsequent transactions.
11. The server of claim 8 , wherein generating the false positive score accounts for whether a channel of the declined transaction and subsequent transactions is the same.
12. The server of claim 8 , wherein generating the false positive score accounts for demographics information of the user for a card making a subsequent transaction and the user of the declined transaction including age, income, and gender.
13. The server of claim 8 , wherein generating the false positive score accounts for the user profile of the user for a card making a subsequent transaction and the profile of the user of the declined transaction.
14. The server of claim 8 , wherein generating the false positive score includes utilizing the following formula:
False Positive Score (FPS)=α+β1*Same/Similar Merchant+β2*Time difference+/β3*Amount difference+/β4*Same/Different channel+β5*similar demographics+β6*similar profile,
where α=fixed component determined from False positive data of issuer, values of coefficients β1, β2, β3, β4, β5, and β 6 are based on assigned significances of predicting variables to predict a probability of a false positive, same/Similar Merchant=0 or 1, same/Different channel=0 or 1, similar demographics=0 or 1, and similar profile=a score between 0 and 1 depending on similarity in profile of a declined card and a card making a next transaction.