IP Library › Granted Patent US 11,263,661
Granted Patent B2
US 11,263,661 · App. 16/895,587 · Granted Mar 1, 2022

Optimal view correction for content

Inventor: Keqing Liang (Sunnyvale, CA)
Assignee: Microsoft Technology Licensing, LLC
G06Q30/0246G06N20/00G06Q10/04G06Q10/1053G06Q30/0249G06Q30/0269G06Q30/0275G06N3/08
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 11,263,661
App. No.
16/895,587
Granted
Mar 1, 2022
Kind
B2
Abstract

In an example embodiment, a bid of an impression of a piece of content, while dynamically set at impression time, may be based on a base bid that is something of a rough indicator of what the estimated price will be. That base bid then is adjusted dynamically up or down at impression time. This base bid can be determined by dividing the expected number of impressions for a day by a total daily budget. The expected number of impressions may be determined by using the empirical number of impressions from the previous day. As such, in an example embodiment, the prediction of the number of impressions for a day utilizes a corrected version of the empirical number of impressions from the prior day, with the corrected version based on a specialized formula with weights trained by a machine learning algorithm.

Claims (42)

1. A system comprising:

a computer-readable medium having instructions stored thereon, which, when executed by a processor, cause the system to perform operations comprising:

training a view forecasting model by passing a plurality of features extracted from training data to a machine learning algorithm, causing the machine learning algorithm to repetitively alter values for one or more weights until a sum of squares of a difference between a view through ratio of the training data and a predicted view through ratio from a portion of the view forecasting model is minimized;

obtaining impression information for a first sponsored piece of content, the impression information including a count of a number of impressions of the first sponsored piece of content displayed in an online network over a first time period;

passing the impression information to a view forecasting model to output a predicted number of impressions of the first sponsored piece of content over a future time period, based on a corrected version of the count of the number of impressions of the first sponsored piece of content over the first time period, the corrected version produced in the view forecasting model by a formula having the one or more weights learned during the training of the view forecasting model, the corrected version being a version in which the count of the number of impressions of the first sponsored piece of content displayed in the online network over the first time period has been fixed;

obtaining a budget for the first sponsored piece of content for the future time period;

calculating a base bid for the first sponsored piece of content based on the budget and the predicted number of impressions of the first sponsored piece of content over the future time period;

determining to display the first sponsored piece of content to a first user;

subsequent to the determining, dynamically updating the base bid based on one or more attributes of the first user, to produce a dynamic bid for the first sponsored piece of content; and

causing the dynamic bid to be deducted from the budget.

2. The system of claim 1 , wherein the calculating the base bid includes dividing the budget by the predicted number of impressions of the first sponsored piece of content over the future time period.

3. The system of claim 2 , wherein the determining to display the first sponsored piece of content to the first user includes passing the one or more features of the first sponsored piece of content and the one or more features of the first user to a successful content model, the successful content model being a machine learned model trained by a machine learning algorithm to output a score indicative of a likelihood that display of the first sponsored piece of content to the first user will result in a positive interaction by the first user with the first sponsored piece of content in a graphical user interface of the online network.

4. The system of claim 3 , wherein the positive interaction is a click.

5. The system of claim 3 , wherein the piece of content is a job listing and the positive interaction is the user applying for a job associated with the job listing through the graphical user interface.

6. The system of claim 3 , wherein the successful content model is a neural network.

7. The system of claim 1 , wherein the operations further comprise displaying the first sponsored piece of content to the first user as an impression in a graphical user interface of the online network.

8. The system of claim 1 , wherein the impression is displayed in a slot designated for sponsored pieces of content on a screen of the graphical user interface, the slot interleaved with one or more slots designated for organic pieces of content.

9. A computerized method comprising:

training a view forecasting model by passing a plurality of features extracted from training data to a machine learning algorithm, causing the machine learning algorithm to repetitively alter values for one or more weights until a sum of squares of a difference between a view through ratio of the training data and a predicted view through ratio from a portion of the view forecasting model is minimized;

obtaining impression information for a first sponsored piece of content, the impression information including a count of a number of impressions of the first sponsored piece of content displayed in an online network over a first time period;

passing the impression information to a view forecasting model to output a predicted number of impressions of the first sponsored piece of content over a future time period, based on a corrected version of the count of the number of impressions of the first sponsored piece of content over the first time period, the corrected version produced in the view forecasting model by a formula having the one or more weights learned during the training of the view forecasting model, the corrected version being a version in which the count of the number of impressions of the first sponsored piece of content displayed in the online network over the first time period has been fixed;

obtaining a budget for the first sponsored piece of content for the future time period;

calculating a base bid for the first sponsored piece of content based on the budget and the predicted number of impressions of the first sponsored piece of content over the future time period;

determining to display the first sponsored piece of content to a first user;

subsequent to the determining, dynamically updating the base bid based on one or more attributes of the first user, to produce a dynamic bid for the first sponsored piece of content; and

causing the dynamic bid to be deducted from the budget.

10. The method of claim 9 , wherein the calculating the base bid includes dividing the budget by the predicted number of impressions of the first sponsored piece of content over the future time period.

11. The method of claim 10 , wherein the determining to display the first sponsored piece of content to the first user includes passing the one or more features of the first sponsored piece of content and the one or more features of the first user to a successful content model, the successful content model being a machine learned model trained by a machine learning algorithm to output a score indicative of a likelihood that display of the first sponsored piece of content to the first user will result in a positive interaction by the first user with the first sponsored piece of content in a graphical user interface of the online network.

12. The method of claim 11 , wherein the positive interaction is a click.

13. The method of claim 11 , wherein the piece of content is a job listing and the positive interaction is the user applying for a job associated with the job listing through the graphical user interface.

14. The method of claim 11 , wherein the successful content model is a neural network.

15. The method of claim 9 , wherein the operations further comprise displaying the first sponsored piece of content to the first user as an impression in a graphical user interface of the online network.

16. The method of claim 9 , wherein the impression is displayed in a slot designated for sponsored pieces of content on a screen of the graphical user interface, the slot interleaved with one or more slots designated for organic pieces of content.

17. A system comprising:

means for training a view forecasting model by passing a plurality of features extracted from training data to a machine learning algorithm, causing the machine learning algorithm to repetitively alter values for one or more weights until a sum of squares of a difference between a view through ratio of the training data and a predicted view through ratio from a portion of the view forecasting model is minimized;

means for obtaining impression information for a first sponsored piece of content, the impression information including a count of a number of impressions of the first sponsored piece of content displayed in an online network over a first time period;

means for passing the impression information to a view forecasting model to output a predicted number of impressions of the first sponsored piece of content over a future time period, based on a corrected version of the count of the number of impressions of the first sponsored piece of content over the first time period, the corrected version produced in the view forecasting model by a formula having the one or more weights learned during the training of the view forecasting model, the corrected version being a version in which the count of the number of impressions of the first sponsored piece of content displayed in the online network over the first time period has been fixed;

means for obtaining a budget for the first sponsored piece of content for the future time period;

means for calculating a base bid for the first sponsored piece of content based on the budget and the predicted number of impressions of the first sponsored piece of content over the future time period;

means for determining to display the first sponsored piece of content to a first user;

means for subsequent to the determining, dynamically updating the base bid based on one or more attributes of the first user, to produce a dynamic bid for the first sponsored piece of content; and

means for causing the dynamic bid to be deducted from the budget.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jun 8, 2020
From: LIANG, KEQING
To: MICROSOFT TECHNOLOGY LICENSING, LLC
Reel/Frame 052867/0881 →
Continuity (2)
Continuation In Part 16232862 · Dec 26, 2018
Related Publication 20200302477A1 · Sep 24, 2020