Systems and methods of making content-based demographics predictions for website cross-reference to related applications

• US 8,412,648 B2
• Filed: 12/21/2009
• Issued: 04/02/2013
• Est. Priority Date: 12/19/2008
• Status: Active Grant

• Current Assignee: IPwe, Inc.
• Original Assignee: Nxntech LLC
• Inventors: Karypis, George, Han, Eui-Hong, Kabbur, Santosh
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): Seck, Ababacar
• Application Number:
  US12/643,916
• Publication Number:
  US 20100223215A1
• Time in Patent Office: 1,198 Days
• Field of Search: 706/12
• US Class Current: 706/12
• CPC Class Codes: G06Q 10/04 Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" market predictions or forecasting for commercial activities G06Q30/0202
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: G06Q 30/0254 based on statistics

Abstract

Systems and methods for making demographic predictions for websites and web-pages. Embodiments include a system and a method of making demographic predictions for websites. The system and method select one or more websites with known demographic attributes for use as training websites, obtain demographic attributes data of the training websites, determine first features of web-pages of the training websites and develop a prediction model using the determined first features and the obtained demographic attributes data. The prediction model predicts one or more values for a target demographic attribute. The system and method determine second features of web-pages of a target website and apply the prediction model to the determined second features of the target website to predict one or more values for the target demographic attribute of the target website.

Claims

1. A method of making demographic predictions for websites the method comprising:

selecting one or more websites with known demographic attributes for use as training websites;

obtaining demographic attributes data of the training websites;

determining first features of web-pages of the training websites;

developing a prediction model using the determined first features and the obtained demographic attributes data, wherein the prediction model predicts one or more values for a target demographic attribute;

determining second features of web-pages of a target website; and

applying the prediction model to the determined second features of the target website to predict one or more values for the target demographic attribute of the target website.

2. The method of claim 1 wherein the developing a prediction model develops the prediction model using support vector regression techniques.

3. The method of claim 1 wherein the determining the first features of web-pages of the training websites includes:

extracting the first features from content of web-pages of the training websites; and

storing representations of the extracted first features in term vectors.

4. The method of claim 3 further comprising:

concatenating the term vectors for each of a subset of web-pages of each of the training websites;

weighting the concatenated term vectors for each of the training websites, wherein developing a prediction model develops the prediction model based on the weighted, concatenated term vectors of each of the training websites.

5. The method of claim 1 wherein the determining the first features of web-pages of the training websites includes determining the linking features of the training websites.

6. The method of claim 1 wherein the determining the first features of web-pages of the training websites includes determining the textual and structural features of web-pages of the training websites.

7. The method of claim 1 wherein the determining the first features of web-pages of the training websites includes determining HTML tag terms of web-pages of the training websites.

8. The method of claim 1 wherein the determining the first features of web-pages of the training websites determines first features from a subset of the web-pages of the training websites.

9. The method of claim 1 wherein developing a prediction model includes developing a prediction model for each of a plurality of discrete values for the target demographic attribute.

10. The method of claim 9 wherein applying the prediction model to the determined second features of the target website includes applying the prediction model for each of the plurality of discrete values of the second demographic attribute to predict discrete values of the target demographic attribute.

11. The method of claim 1 wherein developing a prediction model includes developing a second level prediction model using a cascading learning process.

12. The method of claim 1 wherein developing a prediction model includes developing a distribution prediction model.

38. The method of claim 1 wherein the obtaining demographic attributes data does not obtain browsing behavior or browsing history data of any visitor of the training websites.

13. A method of making demographic predictions for websites, the method comprising:

developing a prediction model using extracted first features and demographic attributes data of a set of training websites, wherein the prediction model predicts one or more values for a target demographic attribute;

determining second features of web-pages of a target website; and

applying the prediction model to the determined second features of the target website to predict one or more values for the target demographic attribute of the target website.

14. The method of claim 13 wherein the determining the second features of web-pages of the target website includes:

extracting the second features from content of web-pages of the training website; and

storing representations of the extracted second features in term vectors.

15. The method of claim 14 further comprising concatenating the term vectors for each of a subset of web-pages of the target website, wherein the applying applies the prediction model to the concatenated term vectors.

16. The method of claim 13 wherein the determining the second features of web-pages of the target website includes determining the linking features of the target website.

17. The method of claim 13 wherein the determining the second features of web-pages of the target website includes determining the textual and structural features of web-pages of the target website.

18. The method of claim 13 wherein the determining the second features of web-pages of the target website includes determining tag terms of web-pages of the target website.

19. The method of claim 14 wherein the determining the second features of web-pages of the target website determines second features from a subset of the web-pages of the target website.

20. The method of claim 13 wherein applying the prediction model to the determined second features of the target website to predict the target demographic attribute of the target website includes applying a prediction model for each of a plurality of discrete values of the target demographic attribute to predict discrete values of the target demographic attribute.

39. The method of claim 13 wherein the demographic attributes data of a set of training websites does not include any browsing behavior or browsing history data of any visitor of the training websites.

21. A method of making demographic predictions for websites comprising:

identifying one or more websites with known demographic characteristics as training websites; and

comparing a first set of features of the training websites to a second set of features of one or more target websites;

wherein the comparing includes;

storing representations of the first set of features in term vectors;

concatenating the term vectors for each of a subset of web-pages of each of the training websites;

weighting the concatenated term vectors for each of the training websites.

22. The method of 21 further defined as correlating the first set of features of the training websites to the known demographic characteristics of the training websites.

25. The method of claim 22 wherein the correlating comprises developing a regression prediction model.

23. The method of 22 further defined as developing a prediction model based upon a correlation of the first set of features of the training websites and the known demographic characteristics of the training websites.

24. The method of 23 further defined as applying the prediction model to the one or more target websites.

26. A method of making demographic predictions for websites comprising:

obtaining demographic attributes data of a set of training websites;

determining first features of web-pages the set of training websites;

correlating the determined first features of web-pages of the set of training websites and the obtained demographic attributes data of the set of training websites;

determining second features of web-pages of a target website; and

predicting values of a second demographic attribute of the target website based on the correlating.

27. The method of claim 26 wherein determining the first features of web-pages of the set of training websites includes:

extracting the first features from content of web-pages of the training websites; and

storing representations of the extracted first features in term vectors.

28. The method of claim 27 wherein the correlating includes using support vector regression techniques.

29. The method of claim 28 wherein the correlating uses support vector regression techniques to generate a prediction model that is used in the predicting.

40. The method of claim 26 wherein the obtaining demographic attributes data does not obtain browsing behavior or browsing history data of any visitor of the training websites.

30. A non-transitory computer readable medium comprising instructions stored thereon that may be executed by a computer for making demographic predictions for websites without using user information for the websites by:

developing a prediction model using extracted first features and demographic attributes data of a set of training websites, wherein the prediction model predicts one or more values for a target demographic attribute;

determining second features of web-pages of a target website; and

applying the prediction model to the determined second features of the target web site to predict one or more values for the target demographic attribute of the target web site.

31. The computer readable medium of claim 30 further including instructions to develop the prediction model using support vector regression techniques.

32. The non-transitory computer readable medium of claim 30 further including instructions to determine the first features of web-pages of the training websites by:

extracting the first features from content of web-pages of the training websites; and

storing representations of the extracted first features in term vectors.

33. The non-transitory computer readable medium of claim 32 further including instructions for:

concatenating the term vectors for each of a subset of web-pages of each of the training websites;

weighting the concatenated term vectors for each of the training websites, wherein developing a prediction model develops the prediction model based on the weighted, concatenated term vectors of each of the training websites.

34. The non-transitory computer readable medium of claim 32 further including instructions to determine first features from a subset of the web-pages of the training websites.

35. The non-transitory computer readable medium of claim 32 further including instructions for developing a prediction model for each of a plurality of discrete values for the target demographic attribute.

36. The non-transitory computer readable medium of claim 32 further including instructions for applying the prediction model to the determined second features of the target website to predict a second demographic attribute of the target website by applying the prediction model for each of the plurality of discrete values of the second demographic attribute to predict discrete values of the second demographic attribute.

41. The non-transitory computer readable medium of claim 30 wherein the demographic attributes data of a set of training websites does not include any browsing behavior or browsing history data of any visitor of the training websites.

37. A system for making demographic predictions for websites, comprising:

a processor;

a memory, wherein the memory includes instructions stored thereon that are executed by the processor for making content-based demographic predictions for websites without using user information for the websites by;

developing a prediction model using extracted first features and demographic attributes data of a set of training websites, wherein the prediction model predicts one or more values for a target demographic attribute;

determining second features of web-pages of the target website; and

applying, utilizing the processor, the prediction model to the determined second features of the target website to predict the target demographic attribute of the target website; and

a network connection for connecting to a network and accessing training and target websites.

42. The system of claim 37 wherein the demographic attributes data of a set of training websites does not include any browsing behavior or browsing history data of any visitor of the training websites.