US 12,014,426 B2
Method of controlling for undesired factors in machine learning models
Jeffrey S. Myers, Normal, IL (US); Kenneth J. Sanchez, San Francisco, CA (US); and Michael L. Bernico, Bloomington, IL (US)
Assigned to State Farm Mutual Automobile Insurance Company, Bloomington, IL (US)
Filed by STATE FARM MUTUAL AUTOMOBILE INSURANCE COMPANY, Bloomington, IL (US)
Filed on Oct. 11, 2022, as Appl. No. 17/963,397.
Application 17/963,397 is a continuation of application No. 16/893,041, filed on Jun. 4, 2020, granted, now 11,501,133.
Application 16/893,041 is a continuation of application No. 15/383,499, filed on Dec. 19, 2016, granted, now 10,769,518, issued on Sep. 8, 2020.
Claims priority of provisional application 62/273,624, filed on Dec. 31, 2015.
Claims priority of provisional application 62/272,184, filed on Dec. 29, 2015.
Prior Publication US 2023/0032355 A1, Feb. 2, 2023
Int. Cl. G06N 3/04 (2023.01); G06N 3/08 (2023.01); G06N 20/00 (2019.01); G06Q 40/08 (2012.01); G06V 10/82 (2022.01); G06V 30/19 (2022.01); G06V 40/16 (2022.01); H04N 7/18 (2006.01); G06Q 30/0207 (2023.01)
CPC G06Q 40/08 (2013.01) [G06N 3/04 (2013.01); G06N 3/08 (2013.01); G06N 20/00 (2019.01); G06V 10/82 (2022.01); G06V 30/19173 (2022.01); G06V 40/169 (2022.01); H04N 7/185 (2013.01); G06Q 30/0207 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method for training and using a machine learning model comprising, via one or more processors:
receiving an unstructured training data set including one or both of images or audio of a plurality of individuals, including individuals having a range of different ages, genders, ethnicities, and races;
training the machine learning model using the unstructured training data set to produce a first trained machine learning model that contains at least one undesired factor;
identifying one or more of the at least one undesired factor, including factors relating to one or more of age, gender, ethnicity, or race, contained in the first trained machine learning model;
training the first trained machine learning model based upon the identified one or more undesired factors to produce a second trained machine learning model trained to identify the identified one or more undesired factors; and
wherein the second trained machine learning model is usable to analyze one or both of images or audio of an insurance applicant for underwriting while excluding the identified one or more undesired factors.