IP Library › Granted Patent US 10,210,548

Granted Patent B1

US 10,210,548 · App. 14/231,362 · Granted Feb 19, 2019

Predictive recommendation system using absolute relevance

Inventor: Lawrence Lee Wai (Palo Alto, CA)

Assignee: GROUPON, INC.

G06Q30/0269 G06Q30/0246

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 10,210,548
App. No.: 14/231,362
Granted: Feb 19, 2019
Kind: B1

External Links

USPTO Patent Center↗ Full Patent PDF↗

Abstract

In general, embodiments of the present invention provide systems, methods and computer readable media for ranking promotions selected for recommendation to consumers based on predictions of promotion performance and consumer behavior. In embodiments, a set of promotions to be recommended to a consumer can be sorted and/or ranked according to respective relevance scores representing a probability that the consumer's behavior in response to the promotion will match a ranking target. In embodiments, calculating scores is based on a relevance model (a predictive function) derived from one or more contextual data sources representing attributes of promotions and consumer behavior. In embodiments, an absolute relevance score represents an absolute prediction of a ranking target variable. In embodiments, absolute relevance may be used to determine personalized local merchant discovery frontiers; featured result set thresholding for impressions; and/or promotion notification triggers. In embodiments, predictive models based on gross revenue may be optimized using promotion category-dependent price boosting.

Claims (104)

1. A computer-implemented method, comprising:

receiving, by a processor, input data representing a request from a consumer device associated with a consumer, the input data including a consumer location;

receiving, by the processor and from a user profile repository, user data describing attributes of the consumer;

receiving, by the processor and from a promotion inventory, promotion data describing attributes of each of a set of available promotions for the consumer;

calculating, by the processor, an initial absolute relevance score for each of the set of promotions using the promotion data and the user data, wherein the initial absolute relevance score represents an estimated absolute conversion probability for the promotion, and wherein the calculating is based on a distance-independent relevance model, wherein the relevance model is a predictive function;

generating, by the processor, a subset of the available promotions by selecting promotions having an initial absolute relevance score above an absolute relevance score threshold;

determining, by the processor, an optimal discovery region size for the consumer using the promotion data describing the subset of the available promotions;

ranking, by the processor, the subset of available promotions based on their respective initial absolute relevance scores; and

presenting the subset of available promotions to the consumer device in accordance with their rankings;

wherein determining the optimal discovery region size comprises:

receiving initialization parameters including a target quantity of promotions to be scored by a core conversion rate model, a minimum conversion rate of interest to the consumer, and a promotion distance radius;

receiving a first subset of the available promotions that each are located within the promotion distance radius of the consumer location;

calculating a first count of the first subset of the available promotions that are associated with an absolute relevance score that is greater than or equal to the minimum conversion rate of interest;

comparing the first count to a scores count threshold; and

in an instance in which the first count is greater than or equal to the scores count threshold,

comparing the first count to the target quantity of promotions; and

in an instance in which the first count is less than or equal to the target quantity,

selecting the first subset of promotions to be scored by the core conversion rate model.

2. The method of claim 1 , further comprising:

calculating, by the processor, featured result set thresholding using the ranked subset of available promotions.

3. The method of claim 1 , further comprising:

generating, by the processor, at least one personalized notification trigger for the consumer using the ranked subset of available promotions, the promotion data, and the user data.

4. The method of claim 1 , further comprising:

in an instance in which the first count is less than the scores count threshold,

increasing, by the processor, the promotion distance radius;

receiving, by the processor, a second subset of the available promotions that each are located within the increased promotion distance radius of the consumer location; and

calculating, by the processor, a second count of the second subset of the available promotions that are associated with an absolute relevance score that is greater than or equal to the minimum conversion rate of interest.

5. The method of claim 1 , further comprising:

in an instance in which the first count is greater than the target quantity,

decreasing, by the processor, the promotion distance radius;

receiving, by the processor, a third subset of the available promotions that each are located within the decreased promotion distance radius of the consumer location; and

calculating, by the processor, a third count of the third subset of the available promotions that are associated with an absolute relevance score that is greater than or equal to the minimum conversion rate of interest.

6. The method of claim 1 , wherein determining an optimal discovery region size for the consumer includes determining a personalized local merchant discovery frontier for the consumer.

7. The method of claim 1 , wherein the predictive function is generated offline using supervised machine learning in a set of modeling stages in which the predictive function is adapted based on training data sets of features that are extracted from a set of data sources.

8. The method of claim 1 , wherein the predictive function is an ensemble of trees and a definition of the predictive function is represented in Predictive Model Markup Language (PMML).

9. The method of claim 7 wherein the predictive function definition is a file comprising one or more Predictive Model Markup Language (PMML) model definitions and a decision tree for use in selecting a predictive model definition from the one or more Predictive Model Markup Language (PMML) model definitions.

10. The method of claim 7 , wherein the file is a JavaScript Object Notation (JASON) file.

11. The method of claim 7 , wherein the predictive function includes multiple random forest models, each representing a different contextual data source.

12. A computer program product, encoded on a computer-readable medium, operable to cause a processor of a data processing apparatus to:

receive input data representing a request from a consumer device associated with a consumer, the input data including a consumer location;

receive, from a user profile repository, user data describing attributes of the consumer;

receive, from a promotion inventory, promotion data describing attributes of each of a set of available promotions for the consumer;

calculate an initial absolute relevance score for each of the set of promotions using the promotion data and the user data, wherein the initial absolute relevance score represents an estimated absolute conversion probability for the promotion, and wherein the calculating is based on a distance-independent relevance model, wherein the relevance model is a predictive function;

generate a subset of the available promotions by selecting promotions having an initial absolute relevance score above an absolute relevance score threshold;

determine an optimal discovery region size for the consumer using the promotion data describing the subset of the available promotions;

rank the subset of available promotions based on their respective initial absolute relevance scores; and

present the subset of available promotions to the consumer device in accordance with their rankings;

wherein determining the optimal discovery region size comprises:

receiving a first subset of the available promotions that each are located within the promotion distance radius of the consumer location;

comparing the first count to a scores count threshold; and

in an instance in which the first count is greater than or equal to the scores count threshold,

comparing the first count to the target quantity of promotions; and

in an instance in which the first count is less than or equal to the target quantity,

selecting the first subset of promotions to be scored by the core conversion rate model.

13. The computer program product of claim 12 , wherein the processor is further caused to:

calculate featured result set thresholding using the ranked subset of available promotions.

14. The computer program product of claim 12 , wherein the processor is further caused to:

generate at least one personalized notification trigger for the consumer using the ranked subset of available promotions, the promotion data, and the user data.

15. The computer program product of claim 12 , wherein the processor is further caused to:

in an instance in which the first count is less than the scores count threshold,

increase the promotion distance radius;

receive a second subset of the available promotions that each are located within the increased promotion distance radius of the consumer location; and

calculate a second count of the second subset of the available promotions that are associated with an absolute relevance score that is greater than or equal to the minimum conversion rate of interest.

16. The computer program product of claim 12 , wherein the processor is further caused to:

in an instance in which the first count is greater than the target quantity,

decrease the promotion distance radius;

receive a third subset of the available promotions that each are located within the decreased promotion distance radius of the consumer location; and

calculate a third count of the third subset of the available promotions that are associated with an absolute relevance score that is greater than or equal to the minimum conversion rate of interest.

17. The computer program product of claim 12 , wherein determining an optimal discovery region size for the consumer includes determining a personalized local merchant discovery frontier for the consumer.

18. An apparatus comprising at least one processor and at least one memory including computer program code, the at least one memory and the computer program code configured to, with the at least one processor, cause the apparatus to: