Determining a model refresh rate

• US 8,782,197 B1
• Filed: 07/17/2012
• Issued: 07/15/2014
• Est. Priority Date: 07/17/2012
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Liu, Jia, Bai, Yijian, Patil, Manojav, Ravichandran, Deepak, Jiampojamarn, Sittichai, Ponnekanti, Shankar
• Primary Examiner(s): JEAN GILLES, JUDE
• Application Number:
  US13/551,236
• Time in Patent Office: 728 Days
• Field of Search: 709/223
• US Class Current: 709/223
• CPC Class Codes: G06Q 30/0272 Period of advertisement exposure
• CPC Class Codes: G06Q 30/0277 Online advertisement

Abstract

A computerized method for determining a model refresh rate for a model representing a list of network user identifiers includes receiving a first and a second model representing the list at a first and a second time, respectively. Similarity between the first and second model is calculated. If the similarity is less than a threshold value, the model refresh rate is a predetermined rate. If the similarity is equal to or exceeds the threshold value, the model refresh rate is decreased to a less frequent rate. A similarity between a reference model and a selected model in a same content category may be calculated. If the similarity is less than a threshold value, the model refresh rate for the selected model is a predetermined rate. If the similarity is equal to or exceeds the threshold value, the model refresh rate for the reference model is applied to the selected model.

Claims

1. A computerized method for determining a model refresh rate for a model representing a list of network user identifiers, the method comprising:

receiving, at a processing circuit, a first model representing the list of network user identifiers at a first time and a second model representing the list of network user identifiers at a second time, wherein the first time is prior to the second time;

calculating, by the processing circuit, a level of similarity between the first model and the second model;

comparing, by the processing circuit, the level of similarity to a threshold value; and

determining, by the processing circuit, the model refresh rate for the list of network user identifiers based in part on a comparison between the level of similarity and the threshold value,

wherein if the level of similarity is less than the threshold value, the model refresh rate is a predetermined rate.

2. The method of claim 1, wherein the first time is a previous day and the second time is a current day.

3. The method of claim 1, wherein the threshold value is less than or equal to 95%.

4. The method of claim 1, wherein if the level of similarity is equal to or exceeds the threshold value, the model refresh rate is decreased to a value less frequent than the predetermined rate.

5. The method of claim 4, wherein the predetermined rate is daily.

6. The method of claim 1, wherein the comparing step is repeated daily until the level of similarity meets or exceeds the threshold value each day for at least a week.

7. A computerized method for determining a model refresh rate for a selected model, the method comprising:

receiving, at a processing circuit, a reference model and a model refresh rate for the reference model, wherein the reference model is an existing version of a model associated with a selected list of network user identifiers;

classifying, by the processing circuit, the reference model based on a predetermined list of content categories;

calculating, by the processing circuit, a level of similarity between the reference model and the selected model, wherein the selected model is one of a plurality of models in a same content category as the reference model and the selected model is an existing version of a model associated a list of network user identifiers, different from the selected list of network user identifiers associated with the reference model;

comparing the level of similarity between the reference model and the selected model to a first threshold value; and

determining, by the processing circuit, the model refresh rate for the selected model based in part on a comparison between the level of similarity between the reference model and the selected model and the first threshold value.

8. The method of claim 7, wherein determining the model refresh rate for the reference model comprises:

receiving, by the processing circuit, a first model representing the reference model at a first time and a second model representing the reference model at a second time, wherein the first time is prior to the second time;

calculating, by the processing circuit, a level of similarity between the first model and the second model;

comparing, by the processing circuit, the level of similarity between the first model and the second model to a second threshold value; and

determining, by the processing circuit, the model refresh rate for the reference model based in part on a comparison between the level of similarity and the threshold value.

9. The method of claim 7, wherein the characteristic data used to classify the reference model comprises vertical interests, list size, or freshness of the list.

10. The method of claim 7, further comprising repeating the steps recited in claim 7 for an other model of the plurality of models in the same content category, the other model different from the selected model and the reference model.

11. The method of claim 7, wherein the first threshold value is less than or equal to 95%.

12. The method of claim 7, wherein the second threshold value is less than or equal to 95%.

13. The method of claim 7, wherein if the level of similarity between the reference model and the selected model is less than the first threshold value, the model refresh rate for the selected model is a predetermined rate, and if the level of similarity between the reference model and the selected model is equal to or exceeds the first threshold value, the model refresh rate for the reference model is used as the model refresh rate for the selected model.

14. The method of claim 13, wherein the predetermined rate is daily.

15. The method of claim 7, wherein the model refresh rate for the selected model is used as a model refresh rate for each of the plurality of models in the same content category.

16. A system for determining a model refresh rate for a selected model, the system comprising a processing circuit operable to:

receive a reference model and a model refresh rate for the reference model, wherein the reference model is an existing version of a model associated with a selected list of network user identifiers;

classify the reference model based on a predetermined list of content categories;

calculate a level of similarity between the reference model and the selected model, wherein the selected model is one of a plurality of models in a same content category as the reference model and the selected model is an existing version of a model associated a list of network user identifiers, different from the selected list of network user identifiers associated with the reference model;

compare the level of similarity between the reference model and the selected model to a first threshold value; and

determine the model refresh rate for the selected model based in part on a comparison between the level of similarity between the reference model and the selected model and the first threshold value.

17. The system of claim 16, wherein to determine the mode refresh rate for the reference model, the processing circuit is further operable to:

receive a first model representing the reference model at a first time and a second model representing the reference model at a second time, wherein the first time is prior to the second time;

calculate a level of similarity between the first model and the second model;

compare the level of similarity between the first model and the second model to a second threshold value; and

determine the model refresh rate for the reference model based in part on a comparison between the level of similarity and the threshold value.

18. The system of claim 16, wherein the characteristic data used to classify the reference model comprises vertical interests, list size, or freshness of the list.

19. The system of claim 16, wherein the first threshold value is less than or equal to 95%.

20. The system of claim 16, wherein the second threshold value is less than or equal to 95%.

21. The system of claim 16, wherein if the level of similarity between the reference model and the selected model is less than the first threshold value, the model refresh rate for the selected model is a predetermined rate, and if the level of similarity between the reference model and the selected model is equal to or exceeds the first threshold value, the model refresh rate for the reference model is used as the model refresh rate for the selected model.

22. The system of claim 21, wherein the predetermined rate is daily.

23. The system of claim 16, wherein the model refresh rate for the selected model is used as a model refresh rate for each of the plurality of models in the same content category.