US 11,682,018 B2
Machine learning model and narrative generator for prohibited transaction detection and compliance
Venkatesh J. Ramesh, Madurai (IN); Rajkumar Baskaran, Vellore (IN); and Swaminathan Raghavan, Chennai (IN)
Assigned to PAYPAL, INC., San Jose, CA (US)
Filed by PAYPAL, INC., San Jose, CA (US)
Filed on Mar. 27, 2020, as Appl. No. 16/833,475.
Prior Publication US 2021/0304204 A1, Sep. 30, 2021
Int. Cl. G06Q 20/40 (2012.01); G06N 3/049 (2023.01); G06N 20/00 (2019.01); G06F 18/21 (2023.01)
CPC G06Q 20/4016 (2013.01) [G06F 18/217 (2023.01); G06N 3/049 (2013.01); G06N 20/00 (2019.01)] 20 Claims
OG exemplary drawing
 
1. A system comprising:
a non-transitory memory; and
one or more hardware processors coupled to the non-transitory memory and configured to read instructions from the non-transitory memory to cause the system to perform operations comprising:
receiving training data comprising transactions indicating a prohibited behavior by one or more accounts associated with the transactions;
training, using a supervised machine learning technique, a machine learning model based on feature data extracted from the training data for features of the machine learning model, wherein the training causes the machine learning model to label the transactions according to one or more trained classifications;
determining, using the features and the machine learning model, a first transaction within the training data that has a first prohibited behavior flag based on the one or more trained classifications;
determining a first descriptive narrative for the first prohibited behavior flag of the first transaction using the features of the first transaction, wherein the first descriptive narrative comprises readable text converted from the features, wherein the determining the first descriptive narrative comprises:
applying a machine learning prediction explainer that identifies a weight of one or more of the features in determining the first prohibited behavior flag from the training of the machine learning model based on the one or more trained classifications, and
generating the readable text from the weight of the one or more of the features and additional explanation data for the first transaction; and
retraining the machine learning model for prohibited transaction pattern identification based on the supervised machine learning technique comprising one of a gradient boosting technique or a random forest technique, the training data, and model training feedback associated with the first descriptive narrative, wherein the machine learning model is configured to perform prohibited transaction identification.