Granted Patent B2

US 12,236,320 · App. 16/866,844 · Granted Feb 25, 2025

Inventors: Peter Joseph Ross (Dublin, IE); Damian Kelly (Kildare, IE)

Assignee: Optum Services (Ireland) Limited

G06N20/00G06F18/214G06N5/04

Abstract

Claims (120)

1. A computer-implemented method comprising:

receiving, by one or more processors, incoming monitoring data associated with a user profile;

generating, by the one or more processors, anomaly-reduced incoming monitoring data by removing values within the incoming monitoring data that are estimated to be mis-recorded based at least in part on a distribution of the incoming monitoring data;

generating, by the one or more processors, an incoming monitoring window based at least in part on the anomaly-reduced incoming monitoring data, wherein the incoming monitoring window comprises a subset of the incoming monitoring data;

accessing, by the one or more processors, a heightened need prediction machine learning model that is trained to determine heightened need prediction for an input incoming monitoring window based at least in part on a heightened need ground-truth that (i) comprises one or more heightened need conditions that are determined based at least in part on training monitoring data that is associated with the user profile and (ii) represents whether an end user associated with the user profile engaged in one or more need-response actions that are (a) associated with one or more target addictive activities or substances and (b) responsive to the one or more heightened need conditions;

generating, by the one or more processors using the heightened need prediction machine learning model, a heightened need prediction for the incoming monitoring window, wherein the heightened need prediction indicates that the incoming monitoring window is associated with a need that is beyond a threshold level of need that is associated with the one or more target addictive activities or substances; and

based at least in part on the heightened need prediction and using an optimal action determination machine learning model that is configured to generate one or more need reduction likelihood values or one or more need reduction speed values for one or more candidate need-reducing actions, initiating, by the one or more processors, performance of a responsive action from the one or more candidate need-reducing actions based at least in part on the one or more need reduction likelihood values or the one or more need reduction speed values, wherein the responsive action comprises communicating with one or more client devices associated with the user profile.

2. The computer-implemented method of claim 1 , wherein the training monitoring data comprises preexisting training monitoring data and training the heightened need prediction machine learning model comprises:

receiving the preexisting training monitoring data associated with the user profile;

generating anomaly-reduced preexisting training monitoring data by performing anomaly reduction on the preexisting training monitoring data;

generating one or more preexisting training monitoring windows based at least in part on the anomaly-reduced preexisting training monitoring data;

for each preexisting training monitoring window of the one or more preexisting training monitoring windows:

generating an offline-inferred heightened need prediction, and

determining an offline-inferred heightened need ground-truth; and

generating the heightened need prediction machine learning model based at least in part on the offline-inferred heightened need prediction and the offline-inferred heightened need ground-truth for each of the one or more preexisting training monitoring windows.

3. The computer-implemented method of claim 2 , wherein the offline-inferred heightened need ground-truth is based at least in part on user monitoring data received from one or more user monitoring client devices associated with the user profile.

4. The computer-implemented method of claim 2 , wherein the offline-inferred heightened need ground-truth represents whether a preexisting training monitoring window is associated with a given need-response action associated with the user profile.

5. The computer-implemented method of claim 1 , wherein the training monitoring data comprises incoming training monitoring data and training the heightened need prediction machine learning model comprises:

receiving the incoming training monitoring data associated with the user profile;

generating anomaly-reduced incoming training monitoring data by performing anomaly reduction on the incoming training monitoring data;

generating an incoming training monitoring window based at least in part on the anomaly-reduced incoming training monitoring data;

generating an inferred heightened need prediction for the incoming training monitoring window;

determining an incoming heightened need ground-truth for the incoming training monitoring window; and

updating the heightened need prediction machine learning model based at least in part on the inferred heightened need prediction and the incoming heightened need ground-truth.

6. The computer-implemented method of claim 5 , wherein the incoming heightened need ground-truth represents whether the incoming training monitoring window is associated with a given heightened need condition for the user profile.

7. The computer-implemented method of claim 1 , wherein the training monitoring data comprises preexisting training monitoring data and incoming training monitoring data, and training the heightened need prediction machine learning model comprises:

receiving the preexisting training monitoring data associated with the user profile;

generating anomaly-reduced preexisting training monitoring data by performing anomaly reduction on the preexisting training monitoring data;

generating one or more preexisting training monitoring windows based at least in part on the anomaly-reduced preexisting training monitoring data;

for each preexisting training monitoring window of the one or more preexisting training monitoring windows:

generating a preexisting inferred heightened need prediction, and

determining a preexisting heightened need ground-truth;

generating the heightened need prediction machine learning model based at least in part on the preexisting inferred heightened need prediction and the preexisting heightened need ground-truth;

receiving the incoming training monitoring data associated with the user profile;

generating anomaly-reduced incoming training monitoring data by performing anomaly reduction on the incoming training monitoring data;

generating an incoming training monitoring window based at least in part on the anomaly-reduced incoming training monitoring data;

generating an inferred heightened need prediction for the incoming training monitoring window;

determining an incoming heightened need ground-truth for the incoming training monitoring window; and

updating the heightened need prediction machine learning model based at least in part on the inferred heightened need prediction and the incoming heightened need ground-truth.

8. The computer-implemented method of claim 1 , wherein initiating the performance of the responsive action comprises:

determining, based at least in part on an optimal action determination model, an optimal need-reducing action for the user profile; and

communicating the optimal need-reducing action to at least one of the one or more client devices.

9. A computing system comprising memory and one or more processors communicatively coupled to the memory, the one or more processors configured to:

receive incoming monitoring data associated with a user profile;

generate anomaly-reduced incoming monitoring data by removing values within the incoming monitoring data that are estimated to be mis-recorded based at least in part ona distribution of the incoming monitoring data;

generate an incoming monitoring window based at least in part on the anomaly-reduced incoming monitoring data, wherein the incoming monitoring window comprises a subset of the incoming monitoring data;

access a heightened need prediction machine learning model that is trained to determine heightened need prediction for an input incoming monitoring window based at least in part on a heightened need ground-truth that (i) comprises one or more heightened need conditions that are determined based at least in part on training monitoring data that is associated with the user profile and (ii) represents whether an end user associated with the user profile engaged in one or more need-response actions that are (a) associated with one or more target addictive activities or substances and (b) responsive to the one or more heightened need conditions;

generate, using the heightened need prediction machine learning model, a heightened need prediction for the incoming monitoring window, wherein the heightened need prediction indicates that the incoming monitoring window is associated with a need that is beyond a threshold level of need that is associated with the one or more target addictive activities or substances; and

based at least in part on the heightened need prediction and using an optimal action determination machine learning model that is configured to generate one or more need reduction likelihood values or one or more need reduction speed values for one or more candidate need-reducing actions, initiate performance of a responsive action, from the one or more candidate need-reducing actions based at least in part on the one or more need reduction likelihood values or the one or more need reduction speed values, wherein the responsive action comprises communicating with one or more client devices associated with the user profile.

10. The computing system of claim 9 , wherein the training monitoring data comprises preexisting training monitoring data and the one or more processors are further configured to train the heightened need prediction machine learning model by:

receiving the preexisting training monitoring data associated with the user profile;

generating anomaly-reduced preexisting training monitoring data by performing anomaly reduction on the preexisting training monitoring data;

generating one or more preexisting training monitoring windows based at least in part on the anomaly-reduced preexisting training monitoring data;

for each preexisting training monitoring window of the one or more preexisting training monitoring windows:

generating an offline-inferred heightened need prediction, and

determining an offline-inferred heightened need ground-truth; and

generating the heightened need prediction machine learning model based at least in part on the offline-inferred heightened need prediction and the offline-inferred heightened need ground-truth for each of the one or more preexisting training monitoring windows.

11. The computing system of claim 10 , wherein the one or more processors are further configured to determine the offline-inferred heightened need ground-truth based at least in part on user monitoring data received from one or more user monitoring client devices associated with the user profile.

12. The computing system of claim 10 , wherein the offline-inferred heightened need ground-truth represents whether a preexisting training monitoring window is associated with a given need-response action associated with the user profile.

13. The computing system of claim 9 , wherein the training monitoring data comprises incoming training monitoring data and the one or more processors are further configured to train the heightened need prediction machine learning model is trained by:

receiving the incoming training monitoring data associated with the user profile;

generating anomaly-reduced incoming training monitoring data by performing anomaly reduction on the incoming training monitoring data;

generating an incoming training monitoring window based at least in part on the anomaly-reduced incoming training monitoring data;

generating an inferred heightened need prediction for the incoming training monitoring window;

determining an incoming heightened need ground-truth for the incoming training monitoring window; and

updating the heightened need prediction machine learning model based at least in part on the inferred heightened need prediction and the incoming heightened need ground-truth.

14. The computing system of claim 13 , wherein the incoming heightened need ground-truth represents whether the incoming training monitoring window is associated with a given heightened need condition for the user profile.

15. The computing system of claim 9 , wherein the training monitoring data comprises preexisting training monitoring data and incoming training monitoring data and the one or more processors are further configured to train the heightened need prediction machine learning model by:

receiving the preexisting training monitoring data associated with the user profile;

generating anomaly-reduced preexisting training monitoring data by performing anomaly reduction on the preexisting training monitoring data;

generating one or more preexisting training monitoring windows based at least in part on the anomaly-reduced preexisting training monitoring data;

for each preexisting training monitoring window of the one or more preexisting training monitoring windows:

generating a preexisting inferred heightened need prediction, and

determining a preexisting heightened need ground-truth;

generating the heightened need prediction machine learning model based at least in part on the preexisting inferred heightened need prediction and the preexisting heightened need ground-truth;

receiving the incoming training monitoring data associated with the user profile;

generating anomaly-reduced incoming training monitoring data by performing anomaly reduction on the incoming training monitoring data;

generating an incoming training monitoring window based at least in part on the anomaly-reduced incoming training monitoring data;

generating an inferred heightened need prediction for the incoming training monitoring window;

determining an incoming heightened need ground-truth for the incoming training monitoring window; and

updating the heightened need prediction machine learning model based at least in part on the inferred heightened need prediction and the incoming heightened need ground-truth.

16. The computing system of claim 9 , wherein the one or more processors are further configured to initiate the performance of the responsive action by:

determining, based at least in part on an optimal action determination model, an optimal need-reducing action for the user profile; and

communicating the optimal need-reducing action to at least one of the one or more client devices.

17. One or more non-transitory computer-readable storage media including instructions that, that when executed by one or more processors, cause the one or more processors to:

receive incoming monitoring data associated with a user profile;

generate anomaly-reduced incoming monitoring data by removing values within the incoming monitoring data that are estimated to be mis-recorded based at least in part on a distribution of the incoming monitoring data;

generate an incoming monitoring window based at least in part on the anomaly-reduced incoming monitoring data, wherein the incoming monitoring window comprises a subset of the incoming monitoring data;

access a heightened need prediction machine learning model that is trained to determine heightened need prediction for an input incoming monitoring window based at least in part on a heightened need ground-truth that (i) comprises one or more heightened need conditions that are determined based at least in part on training monitoring data that is associated with the user profile and (ii) represents whether an end user associated with the user profile engaged in one or more need-response actions that are (a) associated with one or more target addictive activities or substances and (b) responsive to the one or more heightened need conditions;

generate, using the heightened need prediction machine learning model, a heightened need prediction for the incoming monitoring window, wherein the heightened need prediction indicates that the incoming monitoring window is associated with a need that is beyond a threshold level of need that is associated with the one or more target addictive activities or substances; and

based at least in part on the heightened need prediction and using an optimal action determination machine learning model that is configured to generate one or more need reduction likelihood values or one or more need reduction speed values for one or more candidate need-reducing actions, initiate performance of a responsive action, from the one or more candidate need-reducing actions based at least in part on the one or more need reduction likelihood values or the one or more need reduction speed values, wherein the responsive action comprises communicating with one or more client devices associated with the user profile.

18. The one or more non-transitory computer-readable storage media of claim 17 , wherein the training monitoring data comprises preexisting training monitoring data and further including instructions that, that when executed by the one or more processors, cause the one or more processors to train the heightened need prediction machine learning model by:

receiving the preexisting training monitoring data associated with the user profile;

generating anomaly-reduced preexisting training monitoring data by performing anomaly reduction on the preexisting training monitoring data;

generating one or more preexisting training monitoring windows based at least in part on the anomaly-reduced preexisting training monitoring data;

for each preexisting training monitoring window of the one or more preexisting training monitoring windows:

generating an offline-inferred heightened need prediction, and

determining an offline-inferred heightened need ground-truth; and

generating the heightened need prediction machine learning model based at least in part on the offline-inferred heightened need prediction and the offline-inferred heightened need ground-truth for each of the one or more preexisting training monitoring windows.

19. The one or more non-transitory computer-readable storage media of claim 17 , wherein the training monitoring data comprises incoming training monitoring data and further including instructions that, that when executed by the one or more processors, cause the one or more processors to train the heightened need prediction machine learning model by:

receiving the incoming training monitoring data associated with the user profile;

generating anomaly-reduced incoming training monitoring data by performing anomaly reduction on the incoming training monitoring data;

generating an incoming training monitoring window based at least in part on the anomaly-reduced incoming training monitoring data;

generating an inferred heightened need prediction for the incoming training monitoring window;

determining an incoming heightened need ground-truth for the incoming training monitoring window; and

updating the heightened need prediction machine learning model based at least in part on the inferred heightened need prediction and the incoming heightened need ground-truth.

20. The one or more non-transitory computer-readable storage media of claim 17 , wherein the training monitoring data comprises preexisting training monitoring data and incoming training monitoring data and further including instructions that, that when executed by the one or more processors, cause the one or more processors to train the heightened need prediction machine learning model by:

receiving the preexisting training monitoring data associated with the user profile;

generating anomaly-reduced preexisting training monitoring data by performing anomaly reduction on the preexisting training monitoring data;

generating one or more preexisting training monitoring windows based at least in part on the anomaly-reduced preexisting training monitoring data;

for each preexisting training monitoring window of the one or more preexisting training monitoring windows:

generating a preexisting inferred heightened need prediction, and

determining a preexisting heightened need ground-truth;

generating the heightened need prediction machine learning model based at least in part on the preexisting inferred heightened need prediction and the preexisting heightened need ground-truth;

receiving the incoming training monitoring data associated with the user profile;

generating anomaly-reduced incoming training monitoring data by performing anomaly reduction on the incoming training monitoring data;

generating an incoming training monitoring window based at least in part on the anomaly-reduced incoming training monitoring data;

generating an inferred heightened need prediction for the incoming training monitoring window;

determining an incoming heightened need ground-truth for the incoming training monitoring window; and

updating the heightened need prediction machine learning model based at least in part on the inferred heightened need prediction and the incoming heightened need ground-truth.

Assignments (1)

Continuity (1)

References Cited (33)