US 12,353,972 B2
Machine learning techniques using iterative feature refinement routines
Christopher A. Hane, Irvine, CA (US); and Vijay S. Nori, Roswell, GA (US)
Assigned to UnitedHealth Group Incorporated, Minnetonka, MN (US)
Filed by UnitedHealth Group Incorporated, Minnetonka, MN (US)
Filed on Oct. 5, 2021, as Appl. No. 17/449,995.
Prior Publication US 2023/0106667 A1, Apr. 6, 2023
Int. Cl. G06N 20/20 (2019.01); G06F 18/211 (2023.01); G06F 18/2115 (2023.01); G06F 18/245 (2023.01); G06F 18/27 (2023.01)
CPC G06N 20/20 (2019.01) [G06F 18/211 (2023.01); G06F 18/245 (2023.01); G06F 18/2115 (2023.01); G06F 18/27 (2023.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method for generating a predictive output for a predictive input that is associated with a plurality of predictive feature values, the computer-implemented method comprising:
identifying, by one or more processors, a plurality of predictiveness measures corresponding to the plurality of predictive feature values;
determining, by the one or more processors, a decision subset of the plurality of predictive feature values based at least in part on the plurality of predictiveness measures, wherein: (i) the decision subset is initialized to comprise all of the plurality of predictive feature values, (ii) determining the decision subset comprises performing one or more feature refinement routine iterations until a target feature refinement routine iteration in which a feature coverage count of the decision subset satisfies a feature coverage count threshold, and (iii) performing a current feature refinement routine iteration comprises, in response to determining that the feature coverage count of the decision subset fails to satisfy the feature coverage count threshold and no defined stopping conditions have been satisfied: (a) modifying the decision subset to exclude at least one of the plurality of predictive feature values in the decision subset that has a lowest one of the plurality of predictiveness measures that are associated with the decision subset, and (b) performing a subsequent feature refinement routine iteration;
determining, by the one or more processors, the predictive output based at least in part on one or more detected predictive trends for the decision subset; and
performing, by the one or more processors, one or more prediction-based actions based at least in part on the predictive output.