| CPC G06N 20/00 (2019.01) [G06F 16/22 (2019.01)] | 20 Claims |
|
1. A method comprising:
training, by a computer system using a dataset including a plurality of transaction request records, a machine learning algorithm, wherein the training includes:
generating, by the computer system using the machine learning algorithm, individual scores for transaction requests included in the plurality of transaction request records, wherein the individual scores are based on characteristics of the transaction requests, wherein the characteristics include information within the dataset that is descriptive of respective transaction requests;
dividing, by the computer system using one or more of the characteristics, the dataset into a plurality of segments of the dataset, each segment including a portion of the transaction request records that have one or more of the characteristics in common;
receiving, at the computer system, a plurality of constraints, wherein the plurality of constraints defines operational limits for evaluating subsequent transaction requests;
calculating, by the computer system using a linear integer programming algorithm according to the plurality of constraints:
a first decision threshold score for a first segment of the plurality of segments using respective individual scores for transaction request records of the first segment; and
a second decision threshold score, different from the first decision threshold score, for a second segment of the plurality of segments using respective individual scores for transaction request records of the second segment;
receiving, by the computer system, a subsequent transaction request that has not been evaluated;
determining, by the computer system using characteristics of the subsequent transaction request, that the subsequent transaction request corresponds to the first segment; and
evaluating, using the machine learning algorithm and the first decision threshold score, the subsequent transaction request, wherein the evaluating includes:
determining a particular individual score for the subsequent transaction request; and
approving the subsequent transaction request in response to determining that the particular individual score satisfies the first decision threshold score.
|