Modifying artificial intelligence modules of a fraud detection computing system

Abstract

A method includes selecting, for evaluation, an entry regarding a transaction from a current transaction-answer (TA) matrix that includes a prediction field for the entry. The method further includes determining whether the prediction field associated with the entry includes an indication that a predicted fraud answer for the transaction is favorable. When unfavorable, the method further includes obtaining inputted data utilized by a fraud assessment model that includes a plurality of artificial intelligence (AI) modules, obtaining additional data regarding the transaction, augmenting the inputted data with the additional data to produce updated data, utilizing the updated data to produce an updated fraud evaluation answer, and determining whether the updated fraud evaluation answer is favorable. When favorable, the method further includes determining a difference between the inputted and the additional data to produce difference data and updating an AI module of the plurality of AI modules based on the difference data.

What is claimed is: 
     
         1 . A method for execution by a fraud detection computing system, the method comprising:
 selecting an entry from a current transaction-answer (TA) matrix for evaluation by a fraud detection computing system, wherein the entry is regarding a transaction, and wherein the current TA matrix includes a prediction field for each entry;   determining whether the prediction field associated with the entry includes an indication that a predicted fraud answer for the transaction is favorable;   when the predicted fraud answer is unfavorable:
 obtaining inputted data utilized by a fraud assessment model utilized to produce the unfavorable predicted fraud answer, wherein the fraud assessment model includes a plurality of artificial intelligence (AI) modules; 
 obtaining additional data regarding the transaction; 
 augmenting the inputted data with the additional data to produce updated data; 
 executing the fraud assessment model utilizing the updated data produce an updated fraud evaluation answer; 
 determining whether the updated fraud evaluation answer is favorable; 
 when the updated fraud evaluation answer is favorable, determining a difference between the inputted and the additional data to produce difference data; and 
 updating an AI module of the plurality of AI modules based on the difference data. 
   
     
     
         2 . The method of  claim 1 , wherein the current TA matrix further comprises one or more of:
 a transaction identifier field;   an auto answer field;   an agent answer field; and   an actual fraud status field.   
     
     
         3 . The method of  claim 1 , wherein the transaction is for a digital service, and wherein the transaction involves a source computing device, a destination computing device, and a data transaction network. 
     
     
         4 . The method of  claim 3 , wherein the digital service comprises one or more of:
 streaming video;   streaming music;   online software;   online acquisition of a gift card;   utilization of the gift card;   an access of digital information via an online data storage service provider; and   a feature of a social media platform.   
     
     
         5 . The method of  claim 1 , wherein generating the indication that the predicted fraud answer for the transaction is unfavorable comprises one of:
 determining a reject indication of an answer field associated with the transaction corresponds with a non-fraudulent indication of an actual fraud status field associated with the transaction; and   determining an accept indication of the answer field associated with the transaction corresponds with a fraudulent indication of an actual fraud status field associated with the transaction.   
     
     
         6 . The method of  claim 1 , wherein generating the indication that the predicted fraud answer for the transaction is favorable comprises one of:
 determining an accept indication of an answer field associated with the transaction corresponds with a non-fraudulent indication of an actual fraud status field associated with the transaction; and   determining a reject indication of the answer field associated with the transaction corresponds with a fraudulent indication of an actual fraud status field associated with the transaction.   
     
     
         7 . The method of  claim 1  further comprises:
 prior to the selecting the entry, rendering fraud evaluation answers regarding a list of transactions, wherein the list of transactions includes the transaction; 
 generating a transaction-answer matrix regarding the fraud evaluation answers; and 
 updating the transaction-answer matrix based on charge back reports and probability reports to produce the current TA matrix. 
 
     
     
         8 . The method of  claim 7 , wherein a charge back report of the charge back reports comprises:
 a charge back to an account associated with the transaction due to the transaction subsequently being deemed fraudulent.   
     
     
         9 . The method of  claim 7 , wherein a probability report of the probability reports comprises:
 a report regarding an agent answer associated with a corresponding transaction of the list of transactions and a validity of the agent answer.   
     
     
         10 . The method of  claim 7 , wherein the rendering a fraud evaluation answer of the fraud evaluation answers regarding the transaction comprises:
 receiving, by the fraud detection computing system, the transaction for fraud evaluation, wherein the transaction is between a first computing device of a data transactional network associated with the fraud detection computing system and a second computing device of the data transactional network;   generating, by the fraud detection computing system, a plurality of evidence vectors regarding the transaction, wherein an evidence vector of the plurality of evidence vectors is a piece of information regarding one or more of data associated with the first computing device, data associated with the second computing device, data associated with a network that supported the transaction, bad actor data, fraud type data, transaction data, and system use data;   engaging, by the fraud detection computing system, a first group of AI modules of the plurality of AI modules to generate a plurality of scores based on at least some of the plurality of evidence vectors and system fraud tolerances; and   automatically generating, by a multi-module fusion AI module of the plurality of AI modules, a fraud evaluation answer for the transaction based on the plurality of scores.   
     
     
         11 . The method of  claim 10 , wherein the fraud evaluation answer is one of:
 an accept the transaction answer;   a rejection the transaction answer; and   a further review the transaction answer.   
     
     
         12 . The method of  claim 11 , wherein when the fraud evaluation answer is the further review the transaction answer, the method further comprises:
 engaging a second group of AI modules of the plurality of AI modules to produce swarm data regarding the first group of AI modules and the fraud detection computing system;   re-engaging the first group of AI modules to produce an adjusted plurality of scores based on the swarm data and the system fraud tolerances; and   automatically generating, by the multi-module fusion AI module, an updated fraud evaluation answer for the transaction based on the adjusted plurality of scores.   
     
     
         13 . The method of  claim 1 , wherein the updating the AI module includes one or more of:
 adjusting a parameter of the AI module;   modifying data of an evidence vector that is inputted into the AI module;   including a score from another AI module as an input to the AI module; and   modifying a weighting of a score produced by the AI module.   
     
     
         14 . The method of  claim 1  further comprises:
 when the updated fraud evaluation answer is unfavorable, determining the difference in the inputted and the additional data to produce the difference data; and 
 creating a new AI module for inclusion in the fraud assessment model based on the difference data. 
 
     
     
         15 . The method of  claim 1 , wherein the plurality of AI modules include:
 a set of risk assessment AI modules;   a set of evidentiary AI modules; and   a set of swarm processing AI modules.   
     
     
         16 . The method of  claim 1  further comprises:
 when the predicted fraud answer is favorable, determining whether a next entry in the matrix exists; and 
 when the next entry exists, selecting the next entry for evaluation by the fraud detection computing system. 
 
     
     
         17 . The method of  claim 1 , wherein when the predicted fraud answer is unfavorable further comprises:
 reconstructing the fraud assessment model.