US 12,148,048 B2
Framework for transaction categorization personalization
Lei Pei, Sunnyvale, CA (US); Juan Liu, Cupertino, CA (US); Ruobing Lu, Mountain View, CA (US); Ying Sun, Foster City, CA (US); Heather Elizabeth Simpson, Sunnyvale, CA (US); and Nhung Ho, Redwood City, CA (US)
Assigned to Intuit Inc., Mountain View, CA (US)
Filed by Intuit Inc., Mountain View, CA (US)
Filed on Mar. 30, 2021, as Appl. No. 17/218,079.
Prior Publication US 2022/0318925 A1, Oct. 6, 2022
Int. Cl. G06Q 40/12 (2023.01); G06N 5/04 (2023.01); G06N 20/00 (2019.01); G06N 3/084 (2023.01)
CPC G06Q 40/12 (2013.12) [G06N 5/04 (2013.01); G06N 20/00 (2019.01); G06N 3/084 (2013.01)] 14 Claims
OG exemplary drawing
 
1. A method comprising:
receiving a first transaction record corresponding to an entity profile;
selecting, responsive to an entity profile failing to satisfy a usage threshold, a general model as a baseline model from a plurality of machine learning models, wherein:
the general model comprises:
a first transaction model that generates a first transaction vector,
a match model that generates a match score between the first transaction vector and a respective account vector, and
a first account selector that selects a first baseline identifier using the match score, and
the plurality of machine learning models comprises the general model and a custom model;
processing, by the baseline model, the first transaction record to select a first account identifier using the first baseline identifier, wherein the processing comprises:
generating, by the first transaction model, the first transaction vector from the first transaction record, the first transaction vector in a same vector space as a plurality of account vectors,
wherein each of the plurality of account vectors are generated for a respective account of a first plurality of accounts,
generating, by the match model, a corresponding match score between the first transaction vector and each of at least two of the plurality of account vectors to generate a plurality of match scores, and
selecting, by the first account selector using the plurality of match scores, the first baseline identifier for a corresponding account of the first plurality of accounts;
presenting the first account identifier for the first transaction record;
receiving, subsequent to the first transaction record, a second transaction record corresponding to the entity profile;
selecting, responsive to the entity profile satisfying the usage threshold, the custom model as the baseline model from the plurality of machine learning models;
processing, by the baseline model, the second transaction record to select a second account identifier using a second baseline identifier, wherein the processing comprises:
generating, by a second transaction model, a second transaction vector from the second transaction record,
generating, by an adapter model, an adapter model output using the second transaction vector,
wherein the adapter model is customized to a customized chart of accounts of an entity corresponding to the entity profile, and wherein the adapter model output comprises a value for each custom account of at least two custom accounts in the customized chart of accounts, the value indicating whether the corresponding custom account is a category for the second transaction record, and
selecting, by a second account selector, the second baseline identifier according to the adapter model output; and
presenting the second account identifier for the second transaction record.