IDENTIFICATION AND MEASUREMENT OF SOCIAL INFLUENCE AND CORRELATION

• US 20110055132A1
• Filed: 08/26/2009
• Published: 03/03/2011
• Est. Priority Date: 08/26/2009
• Status: Active Grant

• Current Assignee: Yahoo AD Tech LLC
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Ravikumar, Shanmugasundaram, Anagnostopoulos, Aris, Mahdian, Mohammad
• Granted Patent: US 8,401,984 B2
• Time in Patent Office: Days
• US Class Current: 706/46
• CPC Class Codes: G06Q 10/00 Administration; Management

Abstract

Techniques for detecting social influence between users in a set of users with regard to an activity are provided. Data for each user of the set of users is received that includes a time value at which the user became active with regard to the activity, and includes at least one indication of another user in the set of users associated with the user. A first estimate of social correlation in the set of users is determined based on the data. The data is modified. For instance, the data may be modified according to a shuffle test and/or an edge reversal test. A second estimate of social correlation in the set of users is determined based on the modified data. The first estimate is compared to the second estimate to determine a degree of social influence in the set of users.

Claims

1. A method for detecting social influence between users in a set of users with regard to an activity, comprising:

receiving data for each user of the set of users that includes a time value at which the user became active with regard to the activity, and includes at least one indication of another user in the set of users associated with the user;

determining a first estimate of social correlation in the set of users based on the data;

modifying the data;

determining a second estimate of social correlation in the set of users based on the modified data; and

comparing the first estimate to the second estimate to determine a degree of social influence in the set of users.

2. The method of claim 1, wherein said determining a first estimate of social correlation in the set of users based on the data comprises:

determining a first number of users Y_c,t of the set of users for each time t of a plurality of times t that had a number of c associated active users at time t and that became active at time t,

determining a first number of users N_c,t of the set of users for each time t of the plurality of times t that were inactive at time t, had a number of c associated active users at time t, and did not become active at time t, and

generating a first estimate of a coefficient α

using a maximum likelihood logistic regression based on the determined first number of users Y_c,t and the determined first number of users N_c,t, wherein the coefficient α

indicates a degree of social correlation in the set of users;

wherein said determining a second estimate of social correlation in the set of users based on the modified data comprises;

determining a second number of users Y_c,t of the set of users for each time t of the plurality of times t based on the modified data,

determining a second number of users N_c,t of the set of users for each time t of the plurality of times t based on the modified data, and

generating a second estimate of the coefficient α

using the maximum likelihood logistic regression based on the determined second number of users Y_c,t and the determined second number of users N_c,t; and

wherein said comparing the first estimate to the second estimate to determine a degree of social influence between the users in the set of users comprises;

comparing the first estimate of the coefficient α

to the second estimate of the coefficient α

to determine the degree of social influence.

3. The method of claim 2, wherein said generating a first estimate of a coefficient α

using a maximum likelihood logistic regression based on the determined first number of users Y_c,t and the determined first number of users N_c,t comprises;

calculating Y_c for each value of c, where

4. The method of claim 1, wherein said modifying the data comprises:

shuffling time values in the data between users of the set of users.

5. The method of claim 1, wherein said modifying the data comprises:

reversing a direction of each indication of association between users of the set of users in the data.

6. The method of claim 1, wherein said comparing the first estimate to the second estimate to determine a degree of social influence between the users in the set of users comprises:

determining that social influence is present in the set of users if a difference between the first estimate and the second estimate of the coefficient α

is greater than a predetermined threshold.

7. The method of claim 1, wherein said comparing the first estimate to the second estimate to determine a degree of social influence between the users in the set of users comprises:

determining that a relatively greater amount of social influence is present in the set of users if a difference between the first estimate and the second estimate of the coefficient α

is a relatively greater value, and that a relatively lesser amount of social influence is present in the set of users if the difference between the first estimate and the second estimate of the coefficient α

is a relatively lesser value.

8. A system for detecting social influence between users in a set of users with regard to an activity, comprising:

a social correlation determiner that receives data for each user of the set of users that includes a time value at which the user became active with regard to the activity, and includes at least one indication of another user in the set of users associated with the user, the social correlation determiner being configured to determine a first estimate of social correlation in the set of users based on the data;

a data modifier configured to modify the data; and

a social correlation comparator;

the social correlation determiner being configured to receive the modified data, and to determine a second estimate of social correlation in the set of users based on the modified data; and

the social correlation comparator being configured to compare the first estimate to the second estimate to determine a degree of social influence in the set of users.

9. The system of claim 8, wherein the social correlation determiner comprises:

an active user determiner;

an inactive user determiner; and

a logistic regression estimator;

wherein the active user determiner is configured to determine a first number of users Y_c,t of the set of users for each time t of a plurality of times t that had a number of c associated active users at time t and that became active at time t;

wherein the inactive user determiner is configured to determine a first number of users N_c,t of the set of users for each time t of the plurality of times t that were inactive at time t, had a number of c associated active users at time t, and did not become active at time t;

wherein the logical regression estimator is configured to generate a first estimate of a coefficient α

using a maximum likelihood logistic regression based on the determined first number of users Y_c,t and the determined first number of users N_c,t, wherein the coefficient α

indicates a degree of social correlation in the set of users;

wherein the active user determiner is configured to determine a second number of users Y_c,t of the set of users for each time t of the plurality of times t based on the modified data;

wherein the inactive user determiner is configured to determine a second number of users N_c,t of the set of users for each time t of the plurality of times t based on the modified data;

wherein the logical regression estimator is configured to generate a second estimate of the coefficient α

using the maximum likelihood logistic regression based on the determined second number of users Y_c,t and the determined second number of users N_c,t; and

wherein the social correlation comparator is configured to compare the first estimate of the coefficient α

to the second estimate of the coefficient α

to determine the degree of social influence.

10. The system of claim 9, wherein the logical regression estimator comprises:

a first summer configured to calculate Y_c for each value of c, where

11. The system of claim 8, the data modifier comprises:

a time value shuffle module configured to shuffle time values in the data between users of the set of users.

12. The system of claim 8, wherein the data modifier comprises:

an edge reversal module configured to reverse a direction of each indication of association between users of the set of users in the data.

13. The system of claim 8, wherein the social correlation comparator comprises:

a subtractor configured to determine a difference between the first estimate and the second estimate of the coefficient α

; and

a threshold comparator configured to compare the determined difference to a predetermined threshold;

wherein social influence is determined to be present in the set of users if the determined difference is greater than the predetermined threshold.

14. The system of claim 8, wherein the social correlation comparator comprises:

a subtractor configured to determine a difference between the first estimate and the second estimate of the coefficient α

;

wherein the social correlation comparator is configured to indicate a degree of social influence proportional to the determined difference.

15. A computer program product comprising a computer-readable medium having computer program logic recorded thereon for enabling a processor to detect social influence between users in a set of users with regard to an activity, comprising:

first computer program logic means for enabling the processor to receive data for each user of the set of users that includes a time value at which the user became active with regard to the activity, and includes at least one indication of another user in the set of users associated with the user;

second computer program logic means for enabling the processor to determine a first estimate of social correlation in the set of users based on the data;

third computer program logic means for enabling the processor to modify the data;

fourth computer program logic means for enabling the processor to determine a second estimate of social correlation in the set of users based on the modified data; and

fifth computer program logic means for enabling the processor to compare the first estimate to the second estimate to determine a degree of social influence in the set of users.

16. The computer program product of claim 15, wherein said second computer program logic means comprises:

sixth computer program logic means for enabling the processor to determine a first number of users Y_c,t of the set of users for each time t of a plurality of times t that had a number of c associated active users at time t and that became active at time t,

seventh computer program logic means for enabling the processor to determine a first number of users N_c,t of the set of users for each time t of the plurality of times t that were inactive at time t, had a number of c associated active users at time t, and did not become active at time t, and

eighth computer program logic means for enabling the processor to generate a first estimate of a coefficient α

using a maximum likelihood logistic regression based on the determined first number of users Y_c,t and the determined first number of users N_c,t, wherein the coefficient α

indicates a degree of social correlation in the set of users;

wherein said fourth computer program logic means comprises;

ninth computer program logic means for enabling the processor to determine a second number of users Y_c,t of the set of users for each time t of the plurality of times t based on the modified data,

tenth computer program logic means for enabling the processor to determine a second number of users N_c,t of the set of users for each time t of the plurality of times t based on the modified data, and

eleventh computer program logic means for enabling the processor to generate a second estimate of the coefficient α

using the maximum likelihood logistic regression based on the determined second number of users Y_c,t and the determined second number of users N_c,t; and

wherein said fifth computer program logic means comprises;

twelfth computer program logic means for enabling the processor to compare the first estimate of the coefficient α

to the second estimate of the coefficient α

to determine the degree of social influence.

17. The computer program product of claim 16, wherein said eighth computer program logic means comprises:

computer program logic means for enabling the processor to calculate Y_c for each value of c, where

18. The computer program product of claim 15, wherein said third computer program logic means comprises:

computer program logic means for enabling the processor to shuffle time values in the data between users of the set of users.

19. The computer program product of claim 15, wherein said third computer program logic means comprises:

computer program logic means for enabling the processor to reverse a direction of each indication of association between users of the set of users in the data.

20. The computer program product of claim 15, wherein said fifth computer program logic means comprises:

computer program logic means for enabling the processor to determine that social influence is present in the set of users if a difference between the first estimate and the second estimate of the coefficient α

is greater than a predetermined threshold.