US 11,734,607 B2
Data clean-up method for improving predictive model training
Akli Adjaoute, Mill Valley, CA (US)
Assigned to Brighterion, Inc., Purchase, NY (US)
Filed by Brighterion, Inc., Purchase, NY (US)
Filed on Oct. 30, 2020, as Appl. No. 17/85,109.
Application 17/085,109 is a continuation of application No. 16/398,917, filed on Apr. 30, 2019, granted, now 10,846,623.
Application 16/398,917 is a continuation of application No. 14/935,742, filed on Nov. 9, 2015, abandoned.
Application 14/935,742 is a continuation in part of application No. 14/815,934, filed on Jul. 31, 2015, abandoned.
Application 14/815,934 is a continuation in part of application No. 14/815,848, filed on Jul. 31, 2015, abandoned.
Application 14/815,848 is a continuation in part of application No. 14/514,381, filed on Oct. 15, 2014, abandoned.
Application 14/935,742 is a continuation in part of application No. 14/521,667, filed on Oct. 23, 2014, abandoned.
Prior Publication US 2021/0049510 A1, Feb. 18, 2021
This patent is subject to a terminal disclaimer.
Int. Cl. G06Q 40/00 (2023.01); G06N 20/00 (2019.01); G06N 5/04 (2023.01); G06F 16/215 (2019.01)
CPC G06N 20/00 (2019.01) [G06F 16/215 (2019.01); G06N 5/04 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method for training a predictive model, comprising:
receiving, at one or more processors, a record including a plurality of data fields;
determining, via the one or more processors, that a data field of the plurality of data fields includes invalid contents using a data dictionary;
generating a revised record based on the record at least in part by substituting, via the one or more processors, a replacement data value for the invalid contents of the data field based on the corresponding determination of invalidity;
generating, via the one or more processors, a record file comprising a plurality of records including the revised record;
executing, via the one or more processors, a computer learning training algorithm to train the predictive model based on the record file, the predictive model—
comprising a plurality of smart agents and a classification model including one or more of data mining logic, a neural network, case-based-reasoning, clustering, and business rules,
being configured to combine a plurality of scores into a single predictive result.