Method and system for predicting personal preferences

• US 7,877,346 B2
• Filed: 06/06/2007
• Issued: 01/25/2011
• Est. Priority Date: 06/06/2007
• Status: Active Grant

• Current Assignee: Nielsen Consumer LLC (Nielsen Holdings plc)
• Original Assignee: Affinnova, Inc. (RR Donnelley & Sons Company)
• Inventors: Karty, Kevin D.
• Primary Examiner(s): Holmes; Michael B.
• Assistant Examiner(s): Fernandez Rivas; Omar F
• Application Number:
  US11/758,877
• Publication Number:
  US 20080306895A1
• Time in Patent Office: 1,329 Days
• Field of Search: 706/15, 706/18, 706/20, 706/21, 706/45, 706/46, 706/48, 706/62, 705/1, 705/10, 705/12, 707/3, 707/6
• US Class Current: 706/62
• 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

Abstract

The invention provides techniques for building multiple predictive models of individuals'"'"' affinities for attributes of objects and/or services. The accuracies of multiple predictive models are measured and the models are combined based on the measurements, resulting in a more accurate predictive model of individual-specific affinities for attributes of the objects and/or services.

Claims

1. A computer-implemented method of developing a model predictive of preferences of people among decision objects having one or more attributes, the method comprising the steps of:

a. obtaining, at a computer, a data set indicative of characteristics and expressed affinity for attributes of the decision objects pertaining to a plurality of people;

b. obtaining, at the computer, sets of predictive models based at least in part on the data set, wherein a first set comprises predictive models based on segmenting the data set by said demographic characteristics, a second set comprises predictive models based on segmenting the data set by said affinity for the attributes, and a third set comprises a non-segmented predictive model based on the data set;

c. selecting a predictive model from each of the first and second sets for a person based on the person'"'"'s characteristics and the person'"'"'s expressed affinity for the attributes of the decision objects;

d. calculating, by a computer processor, the person'"'"'s predicted affinity for the attributes of the decision objects using each of the three selected models;

e. calculating, by a computer processor, a weighting factor for each of the three selected models; and

f. combining, by a computer processor, the calculated predicted affinities for the attributes of the decision objects based on the weighting factors to produce a predictive model for the person'"'"'s affinity for the attributes of the decision objects.

2. The computer-implemented method of claim 1 wherein the decision object is selected from the group comprising investment products, automobiles and subparts thereof, apparel, ensembles of clothing, shoes, watches, jewelry, toys, consumer electronics, recipes, foods, book or magazine covers, newspaper layouts, packaging, perfumes, scents, color palettes, fabrics, furnishings and interior designs, wallpaper, chairs, tables, lamps, couches, bed spreads, quilts, carpets, decorative objects, utilitarian objects, or parts thereof, silverware, flatware, crystal ware, fixtures, architecture, web sites and web pages, software, graphical user interfaces, software features, cable TV packages, vacation packages, transportation schedules, academic programs or curricula;

advertising materials such as printed advertisements, billboard messages, logos, diet plans, exercise plans, restaurant menus, store layouts, office layouts, direct marketing campaigns (fliers and inserts), e-marketing campaigns, business strategies, speech outlines, sales presentations, election campaigns, story lines, business plans, songs, jingles, business services and personal services.

3. The computer-implemented method of claim 1 wherein the attributes of the decision objects comprise one or more of product SKU, benefits, positioning statements, ingredients, efficacy, side effects, dosage, delivery, material components, size, color, shape, price, material, design, shape, contents, packaging, theme, dimension, aspect ratio, dominant color, background color, color scheme, size of print, presence or absence of pictorial materials, manner of presentation, varying types of content for pictorial materials, and text content.

4. The computer-implemented method of claim 1 wherein the characteristics comprise one or more of demographic characteristics, behavioral characteristics and attitudinal characteristics.

5. The computer-implemented method of claim 4 wherein the characteristics comprise demographic characteristics, and the demographic characteristics comprise one or more of age, sex, nationality, political affiliation, income level, geographic affiliation, religion, race, education, profession, height, physical condition, health condition and weight.

6. The computer-implemented method of claim 4 wherein the characteristics comprise behavioral characteristics, and the behavioral characteristics comprise one or more of product/food/service usage, product/food/service consumption behavior, product ownership, daily activities, non-daily activities, personal habits, social habits or activities, transportation behavior, entertainment consumption behavior.

7. The computer-implemented method of claim 4 wherein the characteristics comprise attitudinal characteristics, and the attitudinal characteristics comprise one or more of beliefs, thought patterns, thought habits, opinions, likes and dislikes, and any combination thereof.

8. The computer-implemented method of claim 4 wherein the one or more of the demographic characteristics, behavioral characteristics and attitudinal characteristics are self-reported.

9. The computer-implemented method of claim 4 wherein the one or more of the demographic characteristics, behavioral characteristics and attitudinal characteristics are derived.

10. The computer-implemented method of claim 1 wherein the weighting factors are based on relationships between the person'"'"'s predicted affinity for the attributes of the decision objects calculated using the respective model and the person'"'"'s expressed affinity for such attributes.

11. The computer-implemented method of claim 10 wherein the relationship between the person'"'"'s predicted affinity for the attributes of the decision objects and the person'"'"'s expressed affinity for such attributes comprises a score reflecting the accuracy of the predicted affinity for the attributes of the decision objects to the person'"'"'s expressed affinity for such attributes.

12. The computer-implemented method of claim 11 wherein the weighting factor attributed to each model is proportional to the score for the respective model divided by the combined scores for all of the selected models.

13. The computer-implemented method of claim 1 wherein the first set of predictive models comprises a hierarchical Bayesian model.

14. The computer-implemented method of claim 1 wherein the second set of predictive models comprises a hierarchical Bayesian model.

15. The computer-implemented method of claim 1 further comprising calculating, by the processor, a subset weighting factor for subsets of the attributes using each of the selected models and combining the calculated predicted affinities for each attribute of the decision objects based on the subset weighting factors.

16. The computer-implemented method of claim 1 wherein the first and second predictive models further comprise aggregated, non-hierarchical, group-specific predictive models.

17. The computer-implemented method of claim 1 further comprising recommending a decision object to the person based at least in part on the predictive model for the person'"'"'s affinity for the attributes of the decision objects.

18. An article of manufacture having computer-readable program portions embodied thereon for developing a model predictive of preferences of people among decision objects having a plurality of attributes, the article comprising computer-readable instructions for:

a. obtaining a data set indicative of characteristics and expressed affinity for attributes of the decision objects pertaining to a plurality of people;

b. obtaining sets of predictive models based at least in part on the data set, wherein a first set comprises predictive models based on segmenting the data set by said demographic characteristics, a second set comprises predictive models based on segmenting the data set by said affinity for the attributes, and a third non-segmented predictive model based on the data set;

c. selecting a predictive model from each of the first and second sets for a person based on the person'"'"'s demographic characteristics and the person'"'"'s expressed affinity for the attributes of the decision objects;

d. calculating the person'"'"'s predicted affinity for the attributes of the decision objects using each of the three selected models;

e. calculating a weighting factor for each of the three selected models; and

f. combining the calculated predicted affinities for the attributes of the decision objects based on the weighting factors to produce a predictive model for the person'"'"'s affinity for the attributes of the decision objects.

19. The article of manufacture of claim 18 wherein the characteristics comprise one or more of demographic characteristics, behavioral characteristics and attitudinal characteristics.

20. The article of manufacture of claim 18 further comprising computer-readable instructions for repeating steps c) through f) for a plurality of people to obtain a plurality of predictive models and aggregating the predictive models into an aggregate model for predicting affinities for attributes of the decision objects among a general population.

21. The article of manufacture of claim 18 further comprising computer-readable instructions for calculating a subset weighting factor for subsets of the attributes of the decision objects using each of the models attributed to the person, and combining the calculated predicted affinities for each attribute of the decision objects based on the subset weighting factors.

22. A system for developing a model predictive of preferences of people among decision objects having a plurality of attributes, the system comprising

a modeling engine configured to:

i) apply sets of predictive models to a data set indicative of characteristics and expressed affinity for attributes of a plurality of decision objects pertaining to a plurality of people, wherein a first set of predictive models comprises predictive models segmented by the characteristics, a second set of predictive models comprises predictive models segmented by the affinity for the attributes of the decision objects and a third non-segmented predictive model based on the data set;

ii) select a predictive model from each of the first and second sets for a person based on the person'"'"'s characteristics and the person'"'"'s expressed affinity for the attributes of the decision objects;

iii) calculate the person'"'"'s predicted affinity for the attributes of the plurality of decision objects using each of the three selected models;

iv) calculate a weighting factor for each of the three selected models; and

v) combine the calculated predicted affinities for each attribute of the decision objects based on the weighting factors to produce a predictive model for the individual'"'"'s affinity for the attributes of the decision objects.

23. The system of claim 22 further comprising a data storage module for storing the data set.

24. The system of claim 22 wherein the characteristics comprise one or more of demographic characteristics, behavioral characteristics and attitudinal characteristics.

25. The system of claim 22 wherein the modeling engine is further configured to aggregate a plurality of the predictive models attributed to the plurality of people into an aggregate model for predicting affinities for attributes of the decision objects among the plurality of individuals.

26. The system of claim 22 wherein the modeling engine comprises a plurality of processing components, each component configured to perform functions i) through iv) for each of the selected models simultaneously.

27. The system of claim 22 wherein the decision object is selected from the group comprising investment products, automobiles and subparts thereof, apparel, ensembles of clothing, shoes, watches, jewelry, toys, consumer electronics, recipes, foods, book or magazine covers, newspaper layouts, packaging, perfumes, scents, color palettes, fabrics, furnishings and interior designs, wallpaper, chairs, tables, lamps, couches, bed spreads, quilts, carpets, decorative objects, utilitarian objects, or parts thereof, silverware, flatware, crystal ware, fixtures, architecture, web sites and web pages, software, graphical user interfaces, software features, cable TV packages, vacation packages, transportation schedules, academic programs or curricula;

advertising materials such as printed advertisements, billboard messages, logos, diet plans, exercise plans, restaurant menus, store layouts, office layouts, direct marketing campaigns (fliers and inserts), e-marketing campaigns, business strategies, speech outlines, sales presentations, election campaigns, story lines, business plans, songs, jingles, business services and personal services.

28. The method system of claim 22 wherein the attributes of the decision objects comprise one or more of product SKU, benefits, positioning statements, ingredients, efficacy, side effects, dosage, delivery, material components, size, color, shape, price, material, design, shape, contents, packaging, theme, dimension, aspect ratio, dominant color, background color, color scheme, size of print, presence or absence of pictorial materials, manner of presentation, varying types of content for pictorial materials, and text content.

29. The system of claim 22 wherein the characteristics comprise one or more of demographic characteristics, behavioral characteristics and attitudinal characteristics.

30. The system of claim 29 wherein the characteristics comprise demographic characteristics, and the demographic characteristics comprise one or more of age, sex, nationality, political affiliation, income level, geographic affiliation, religion, race, education, profession, height, physical condition, health condition and weight.

31. The system of claim 29 wherein the characteristics comprise behavioral characteristics, and the behavioral characteristics comprise one or more of product/food/service usage, product/food/service consumption behavior, product ownership, daily activities, non-daily activities, personal habits, social habits or activities, transportation behavior, entertainment consumption behavior.

32. The system of claim 29 wherein the characteristics comprise attitudinal characteristics, and the attitudinal characteristics comprise one or more of beliefs, thought patterns, thought habits, opinions, likes and dislikes, and any combination thereof.

33. The system of claim 29 wherein the one or more of the demographic characteristics, behavioral characteristics and attitudinal characteristics are self-reported.

34. The system of claim 29 wherein the one or more of the demographic characteristics, behavioral characteristics and attitudinal characteristics are derived.

35. The method system of claim 22 wherein the weighting factors are based on relationships between the person'"'"'s predicted affinity for the attributes of the decision objects calculated using the respective model and the person'"'"'s expressed affinity for such attributes.

36. The system of claim 35 wherein the relationship between the person'"'"'s predicted affinity for the attributes of the decision objects and the person'"'"'s expressed affinity for such attributes comprises a score reflecting the accuracy of the predicted affinity for the attributes of the decision objects to the person'"'"'s expressed affinity for such attributes.

37. The system of claim 36 wherein the weighting factor attributed to each model is proportional to the score for the respective model divided by the combined scores for all of the selected models.

38. The system of claim 22 wherein the first set of predictive models comprises a hierarchical Bayesian model.

39. The system of claim 22 wherein the second set of predictive models comprises a hierarchical Bayesian model.

40. The system of claim 22 wherein the modeling engine is further configured to calculate a subset weighting factor for subsets of the attributes using each of the selected models and combining the calculated predicted affinities for each attribute of the decision objects based on the subset weighting factors.

41. The system of claim 22 wherein the first and second predictive models further comprise aggregated, non-hierarchical, group-specific predictive models.

42. The system of claim 22 wherein the modeling engine is further configured to recommend a decision object to the person based at least in part on the predictive model for the person'"'"'s affinity for the attributes of the decision objects.