US 11,935,080 B2
Reinforcement machine learning for personalized intelligent alerting
John Bevil Bates, Highland, UT (US); Ryan Elliott Cobourn, Draper, UT (US); Benjamin Russell Gaines, Highland, UT (US); and Brooke Suzanne Wyckoff, Lehi, UT (US)
Assigned to Adobe Inc., San Jose, CA (US)
Filed by ADOBE INC., San Jose, CA (US)
Filed on Mar. 12, 2020, as Appl. No. 16/816,730.
Application 16/816,730 is a continuation of application No. 14/861,772, filed on Sep. 22, 2015, granted, now 10,621,602.
Prior Publication US 2020/0211039 A1, Jul. 2, 2020
Int. Cl. G06Q 30/0204 (2023.01); G06F 16/951 (2019.01); G06Q 10/067 (2023.01); G06N 20/00 (2019.01)
CPC G06Q 30/0204 (2013.01) [G06F 16/951 (2019.01); G06Q 10/067 (2013.01); G06N 20/00 (2019.01)] 19 Claims
OG exemplary drawing
 
1. A computer-implemented method for providing intelligent alerting, the computer-implemented method comprising:
determining an importance of a plurality of metrics by utilizing a reinforcement machine learning model having a machine learning algorithm that learns user behavior by automatically analyzing, via a hardware processor, a consumption pattern of a user, the consumption p attern corresponding to frequency, recency, and data query patterns of the user interacting with an analytics engine according to an on-demand game, the on-demand game including:
choices, corresponding to the plurality of metrics, from a data duel, wherein a choice of a metric indicates that the metric is more critical to the user than another metric of the data duel; and
application programming interface requests associated with the user interacting with the analytics engine;
learning, via the reinforcement machine learning model, similar user behavior of similar users by performing, by the hardware processor, unsupervised machine learning clustering based upon attributes that provide a maximum entropy and based on consumption patterns of the similar users, the similar user behavior being associated with the plurality of metrics;
based on determining the importance of the plurality of metrics and learning the similar user behavior, providing, using the hardware processor, suggested alerts to the user;
receiving, by the hardware processor and from the user, an indication of selected alerts indicating which of the suggested alerts the user prefers and employing the reinforcement machine learning model to leverage user behavior feedback based on receiving the indication of the selected alerts;
combining or deduplicating, by the hardware processor, related alerts of the selected alerts, the related alerts representing a same macro event;
determining, by the hardware processor, that a particular metric of the plurality of metrics has changed in accordance with a threshold of one or more alerts of the selected alerts;
communicating, using the hardware processor, to the user that the particular metric has changed;
determining, by the hardware processor, a statistical significance of respective dimensions associated with the plurality of metrics to rank contributing factors causing the change; and
providing, by the hardware processor, to the user in real time, contextual analysis based on the statistical significance of the respective dimensions.