Reinforcement machine learning for personalized intelligent alerting

• US 10,621,602 B2
• Filed: 09/22/2015
• Issued: 04/14/2020
• Est. Priority Date: 09/22/2015
• Status: Active Grant

• Current Assignee: Adobe Inc.
• Original Assignee: Adobe Inc.
• Inventors: Bates, John Bevil, Cobourn, Ryan Elliott, Gaines, Benjamin Russell, Wyckoff, Brooke Suzanne
• Primary Examiner(s): Rotaru, Octavian
• Application Number:
  US14/861,772
• Publication Number:
  US 20170083929A1
• Time in Patent Office: 1,666 Days
• Field of Search: None
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06Q 10/067 Enterprise or organisation modelling
• CPC Class Codes: G06Q 30/0204 Market segmentation

Abstract

Embodiments of the present invention relate to providing intelligent alerting and automation for marketing analytics software. In implementation, intelligent alerting is initiated by a user, which enables deep learning models to analyze various data patterns. Intelligent alerting learns about preferences and data consumption patterns of the user with marketing analytics software. Intelligent alerting also accounts for and learns from any manually created alerts set up by the user and/or alerts created automatically by anomaly detection. Intelligent alerting analyzes all other users within the organization of the user to find similar users based on their consumption patterns. An on-demand game may be provided to the user to determine the criticality of one metric change over another. This enables intelligent alerting to automatically provide alerts which pass a critical threshold of importance to the user and context to help the user understand why a metric changes in a significant way.

Claims

1. One or more non-transitory computer storage media storing computer-useable instructions that, when executed by at least one hardware processor, cause the at least one hardware processor to:

receive an indication to activate intelligent alerting, the intelligent alerting including activating a machine learning model to;

analyze a usage pattern of marketing analytics software collected from a user, or analyze other users to identify similar users based on usage patterns of the marketing analytics software collected from the other users, the analyzing the usage pattern includes identifying a pattern of recency or frequency of the user interacting with the marketing analytics software, the analyzing the other users includes identifying patterns in how the other users interact with the marketing analytics software;

query each collected usage pattern, and analyze each usage pattern dimension for statistical significance to determine an anomaly and generate context comprising a ranked list of statistical significant contributing factors to the anomaly;

suggest an alert based on the determined anomaly;

determine respective capabilities of first and second user devices based on respective types of the first and second devices;

automatically display to a graphical user interface of the first user device, the context indicating the anomaly in a first format compatible with the capabilities of the first user device;

display to the second device more context or less context than the context displayed to the first user device based on the capabilities of the second device, such that if the second device has a larger display compared to the first device, the context is displayed automatically on the second device, and if the second device has a smaller display compared to the first device, the context is provided to the second device only upon the alert being selected by the user;

determine related alerts with respect to the selected alert and determine other alerts that are no longer relevant to the user; and

deduplicate the related alerts into a single macro event and pause the other alerts that are no longer relevant to the user in order to preserve bandwidth and decrease alert fatigue.

2. The one or more non-transitory computer storage media of claim 1, wherein the stored computer-useable instructions are further executed by the at least one hardware processor to further determine that the user reviews a report at a particular rate of the usage pattern indicating an importance of the report.

3. The one or more non-transitory computer storage media of claim 1, wherein the stored computer-useable instructions are further executed by the at least one hardware processor to further analyze data queries of the marketing analytics software for combinations of segments, time granularity, metrics, dimensions, and dimension items to determine relevance of different data artifact combinations for the user.

4. The one or more non-transitory computer storage media of claim 1, wherein analyzing the other users to identify the similar users based on the usage patterns of the marketing analytics software for the other users comprises utilizing clustering models, decision trees, differences between the other users, and propensity models.

5. The one or more non-transitory computer storage media of claim 1, wherein the stored computer-useable instructions are further executed by the at least one hardware processor to further:

determine that alert will be communicated to an executive user;

prior to communicating the alert to the executive user, receive an approval indication from an analyst; and

upon receiving the approval indication from the analyst, communicate the alert to the executive user.

6. The one or more non-transitory computer storage media of claim 1, wherein the stored computer-useable instructions are further executed by the at least one hardware processor to further determine the anomaly by utilizing deep learning models.