USER ACTION PREDICTION

• US 20170068899A1
• Filed: 09/09/2015
• Published: 03/09/2017
• Est. Priority Date: 09/09/2015
• Status: Active Grant

• Current Assignee: Yahoo Assets LLC
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Laptev, Nikolay Pavlovich, Shu, Xiaokui
• Granted Patent: US 10,643,141 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/954 Navigation, e.g. using categorised browsing
• CPC Class Codes: G06F 16/955 using information identifiers, e.g. uniform resource locators [URL]
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

As provided herein, a webpage navigation of a user over a timeframe and a second webpage navigation of a second user over a second timeframe may be received. A time-variant variable-order Markov model, comprising a context tree, may be generated utilizing the webpage navigation and the second webpage navigation. A third webpage navigation of a third user may be received. A probability that the third user may interact with content, that the third user is a non-human entity, and/or that the third user will access a website may be determined based upon an evaluation of the third webpage navigation using the time-variant variable-order Markov model. A second client device is instructed to present the content to the third user, to present a human verification mechanism to the third user, and/or to instruct a server, providing the website, to alter a server capacity for the website.

Claims

1. A client device for user action prediction, the client device comprising:

a processor;

a display; and

a memory storing instructions that, when executed on the processor, provide a prediction component configured to;

receive a webpage navigation of a user over a timeframe;

receive a second webpage navigation of a second user over a second timeframe, the second timeframe overlapping the timeframe;

generate a time-variant variable-order model, comprising a context tree, utilizing the webpage navigation and the second webpage navigation;

receive a third webpage navigation of a third user;

determine a probability that the third user will interact with content based upon an evaluation of the third webpage navigation using the time-variant variable-order model; and

responsive to the probability exceeding a threshold, instruct a second client device to present the content to the third user.

2. The client device of claim 1, the time-variant variable-order model comprising a Markov model.

3. The client device of claim 1, the prediction component configured to:

receive an extraction of the webpage navigation of the user;

transform the webpage navigation into a format; and

load the webpage navigation, according to the format, into the time-variant variable-order model.

4. The client device of claim 1, the prediction component configured to:

receive an extraction of the second webpage navigation of the second user;

transform the second webpage navigation into a format; and

load the second webpage navigation into the time-variant variable-order model.

5. The client device of claim 1, the prediction component configured to generate the time-variant variable-order model utilizing:

in window aggregation, the in window aggregation outputting;

a webpage location of the user, based upon the webpage navigation, for one or more time points; and

a second webpage location of the second user, based upon the second webpage navigation, for one or more second time points.

6. The client device of claim 1, the time-variant variable-order model configured to at least one of:

identify a first trend indicating the third user has a decreased probability of interacting with the content;

or

identify a second trend indicating the third user has an increased probability of interacting with the content.

7. The client device of claim 1, the prediction component configured to:

prune a branch of the context tree based upon the branch corresponding to an irrelevant node.

8. A method for user action prediction, comprising:

receiving a webpage navigation of a user over a timeframe;

receiving a second webpage navigation of a second user over a second timeframe, the second timeframe overlapping the timeframe;

generating a time-variant variable-order model, comprising a context tree, utilizing the webpage navigation and the second webpage navigation;

receiving a third webpage navigation of a third user;

determining a probability that the third user is a non-human entity based upon an evaluation the third webpage navigation using the time-variant variable-order model; and

responsive to the probability exceeding a threshold, identifying the third user as the non-human entity.

9. The model of claim 8, the time-variant variable-order model comprising a Markov model.

10. The method of claim 8, comprising:

receiving an extraction of the webpage navigation of the user;

transforming the webpage navigation into a format; and

loading the webpage navigation, according to the format, into the time-variant variable-order model.

11. The method of claim 8, comprising:

receiving an extraction of the second webpage navigation of the second user;

transforming the second webpage navigation into a format; and

loading the second webpage navigation, according to the format, into the time-variant variable-order model.

12. The method of claim 8, comprising:

utilizing in window aggregation to output;

a webpage location of the user, based upon the webpage navigation, for one or more time points; and

a second webpage location of the user, based upon the second webpage navigation, for one or more second time points.

13. The method of claim 8, comprising:

pruning a branch of the context tree based upon the branch corresponding to an irrelevant node.

14. The method of claim 8, comprising:

responsive to identifying the third user as the non-human entity, presenting the third user with a human verification mechanism.

15. A client device for user action prediction, the client device comprising:

a processor;

a display; and

a memory storing instructions that, when executed on the processor, provide a prediction component configured to;

receive a webpage navigation of a user over a timeframe;

receive a second webpage navigation of a second user over a second timeframe, the second timeframe overlapping the timeframe;

generate a time-variant variable-order model, comprising a context tree, utilizing the webpage navigation and the second webpage navigation;

receive a third webpage navigation of a third user;

determine a probability that the third user will interact with a website based upon an evaluation of the third webpage navigation using the time-variant variable-order model; and

responsive to the probability exceeding a threshold, instruct a server, providing the website, to alter a server capacity for the website.

16. The client device of claim 15, the time-variant variable-order model comprising a Markov model.

17. The client device of claim 15, the prediction component configured to:

prune a branch of the context tree based upon the branch corresponding to an irrelevant node.

18. The client device of claim 15, the prediction component configured to:

receive an extraction of the webpage navigation of the user;

transform the webpage navigation into a format;

load the webpage navigation, according to the format, into the time-variant variable-order model;

receive an extraction of the second webpage navigation of the second user;

transform the second webpage navigation into the format; and

load the second webpage navigation, according to the format, into the time-variant variable-order model.

19. The client device of claim 15, the prediction component configured to generate the time-variant variable-order model utilizing:

in window aggregation, the in window aggregation outputting;

a webpage location of the user, based upon the webpage navigation, for one or more time points; and

a second webpage location of the user, based upon the second webpage navigation, for one or more second time points.

20. The client device of claim 15, the time-variant variable-order model configured to at least one of:

identify a first trend indicating the third user has a decreased probability of interacting with the website;

or

identify a second trend indicating the third user has an increased probability of interacting with the website.