Methods and systems for categorizing a customer of a service as a churner of a non-churner

1. A method for categorizing a user of a communication service as a churner or non-churner, the method comprising:

obtaining first call data for the user, the first call data identifying communications form the user that were made during a first period of time;

obtaining second call data for the user, the second call data identifying communications from the user that were made during a second period of time;

using the first call data to determine a first value (v 1 ) for a first graph parameter for the user;

using the second call data to determine a second value (v 2 ) for the first graph parameter;

using the first call data to determine a third value (v 3 ) for a second graph parameter for the user, the second graph parameter different from the first graph parameter;

using the second call data to determine a fourth value (v 4 ) for the second graph parameter;

using the determined values v 1 , v 2 , v 3 , and v 4 to determine a churn value (cv); and

categorizing the user as a churner or a non-churner based on said churn value.

2. The method of claim 1 , wherein the churn value is determined by calculating c 1 *(v 2 −v 1 ), wherein c 1 is a predetermined constant.

3. The method of claim 1 ,

wherein the first graph parameter is one of an (a) out-degree parameter, (b) Shapley Value parameter, (c) a proximity prestige parameter, and (d) closeness centrality parameter, and

wherein the second graph parameter being one of: (a) an out-degree parameter, (b) a Shapley Value parameter, (c) a proximity prestige parameter, and (d) closeness centrality parameter.

4. The method of claim 3 , wherein the churn value is determined by calculating c 1 *(v 2 −v 1 ) and c 2 *(v 4 −v 3 ), wherein c 2 is a predetermined constant.

5. The method of claim 3 , further comprising:

using the first call data determining a fifth value (v 5 ) for a third graph parameter for the user, the third graph parameter being one of: (a) an out-degree parameter, (b) a Shapley Value parameter, (c) a proximity prestige parameter, and (d) closeness centrality parameter; and

using the second call data determining a sixth value (v 6 ) for the third graph parameter, wherein

the churn value is determined using v 1 , v 2 , v 3 , v 4 , v 5 and v 6 , and

the third graph parameter is different than the first graph parameter and the second graph parameter.

6. The method of claim 5 , wherein the churn value is determined by calculating c 1 *(v 2 −v 1 ), c 2 *(v 4 −v 3 ), and c 3 *(v 6 −v 5 ), wherein c 3 is a predetermined constant.

7. The method of claim 5 , further comprising:

using the first call data determining a seventh value (v 7 ) for a fourth graph parameter for the user, the fourth graph parameter being one of: (a) an out-degree parameter, (b) a Shapley Value parameter, (c) a proximity prestige parameter, and (d) closeness centrality parameter; and

using the second call data determining an eighth value (v 8 ) for the fourth graph parameter, wherein

the churn value is determined using v 1 , v 2 , v 3 , v 4 , v 5 , v 6 , v 7 and v 8 and

the fourth graph parameter is different than the first graph parameter, the second graph parameter, and the third graph parameter.

8. The method of claim 7 , wherein the churn value is determined by calculating c 1 *(v 2 −v 1 ), c 2 *(v 4 −v 3 ), c 3 *(v 6 −v 5 ), c 4 *(v 8 −v 7 ) wherein c 4 is a predetermined constant.

9. The method of claim 7 , wherein the churn value is determined by calculating c 5 +c 1 *(v 2 −v 1 )+c 2 *(v 4 −v 3 )+c 3 *(v 6 −v 5 )+c 4 *(v 8 −v 7 ), wherein c 5 is a predetermined constant.

10. The method of claim 2 , wherein

the first period of time is first one week period of time, and

the second period of time is second one week period of time that immediately follows the first one week period of time.

11. A churn detector, the churn detector being configured to:

obtain first call data for a user, the first call data identifying communications form the user that were made during a first period of time;

obtain second call data for the user, the second call data identifying communications from the user that were made during a second period of time;

use the first call data to determine a first value (v 1 ) for a first graph parameter for the user;

use the second call data to determine a second value (v 2 ) for the first graph parameter;

use the first call data to determine a third value (v 3 ) for a second graph parameter for the user, the second graph parameter different from the first graph parameter;

use the second call data to determine a fourth value (v 4 ) for the second graph parameter;

use the determined values v 1 , v 2 , v 3 , and v 4 to determine a churn value (cv); and

categorize the user as a churner or a non-churner based on said churn value.

12. The churn detector of claim 11 , wherein the churn value is determined by calculating c 1 *(v 2 −v 1 ), wherein c 1 is a predetermined constant.

13. The churn detector of claim 11 , wherein

the first graph parameter is one of an (a) out-degree parameter, (b) Shapley Value parameter, (c) a proximity prestige parameter, and (d) closeness centrality parameter, and

wherein the second graph parameter being one of: (a) an out-degree parameter, (b) a Shapley Value parameter, (c) a proximity prestige parameter, and (d) closeness centrality parameter.

14. The churn detector of claim 13 , wherein the churn value is determined by calculating c 1 *(v 2 −v 1 ) and c 2 *(v 4 −v 3 ), wherein c 2 is a predetermined constant.

15. The churn detector of claim 13 , wherein the churn detector is further configured to:

use the first call data determining a fifth value (v 5 ) for a third graph parameter for the user, the third graph parameter being one of: (a) an out-degree parameter, (b) a Shapley Value parameter, (c) a proximity prestige parameter, and (d) closeness centrality parameter; and

use the second call data determining a sixth value (v 6 ) for the third graph parameter, wherein

the churn value is determined using v 1 , v 2 , v 3 , v 4 , v 5 and v 6 , and

the third graph parameter is different than the first graph parameter and the second graph parameter.

16. The churn detector of claim 15 , wherein the churn value is determined by calculating c 1 *(v 2 −v 1 ), c 2 *(v 4 −v 3 ), and c 3 *(v 6 −v 5 ), wherein c 3 is a predetermined constant.

17. The churn detector of claim 15 , wherein the churn detector is further configured to:

use the first call data determining a seventh value (v 7 ) for a fourth graph parameter for the user, the fourth graph parameter being one of: (a) an out-degree parameter, (b) a Shapley Value parameter, (c) a proximity prestige parameter, and (d) closeness centrality parameter; and

use the second call data determining an eighth value (v 8 ) for the fourth graph parameter, wherein

the churn value is determined using v 1 , v 2 , v 3 , v 4 , v 5 , v 6 , v 7 and v 8 and

the fourth graph parameter is different than the first graph parameter, the second graph parameter, and the third graph parameter.

18. The churn detector of claim 17 , wherein the churn value is determined by calculating c 1 *(v 2 −v 1 ), c 2 *(v 4 −v 3 ), c 3 *(v 6 −v 5 ), c 4 *(v 8 −v 7 ) wherein c 4 is a predetermined constant.

19. The churn detector of claim 17 , wherein the churn value is determined by calculating c 5 +c 1 *(v 2 −v 1 )+c 2 *(v 4 −v 3 )+c 3 *(v 6 −v 5 )+c 4 *(v 8 −v 7 ), wherein c 5 is a predetermined constant.

20. The churn detector of claim 11 , wherein

the first period of time is first one week period of time, and

the second period of time is second one week period of time that immediately follows the first one week period of time.

21. A computer program product comprising a non-transitory computer readable medium storing a computer code for categorizing a user of a communication service as a churner or non-churner, the computer code comprising:

computer code for obtaining first call data for the user, the first call data identifying communications form the user that were made during a first period of time;

computer code for obtaining second call data for the user, the second call data identifying communications from the user that were made during a second period of time;

computer code for using the first call data to determine a first value (v 1 ) for a graph parameter for the user;

computer code for using the second call data to determine a second value (v 2 ) for the first graph parameter;

computer code for using the first call data to determine a third value (v 3 ) for a second graph parameter for the user, the second graph parameter different from the first graph parameter;

computer code for using the second call data to determine a fourth value (v 4 ) for the second graph parameter;

computer code for using the determined values v 1 , v 2 , v 3 , and v 4 to determine a churn value (cv); and

computer code for categorizing the user as a churner or a non-churner based on said churn value.

22. The computer program product of claim 21 , wherein the computer code for determining the churn value comprises computer code for calculating c 1 *(v 2 −v 1 ), wherein c 1 is a predetermined constant.

23. The method according to claim 1 , further comprising:

using a discriminating analysis model to determine whether the determined value (v 1 ) discriminates between churners and non-churners, wherein

the determined value (v 1 ) is used to determine the churn value (cv) if the determined value (v 1 ) discriminates between churners and non-churners.

24. The method according to claim 3 , wherein the out-degree parameter measures a number of outgoing connections from the user.

25. The method according to claim 1 , wherein the Shapley Value parameter is defined as:

SV

i

=

∑

v

j

∈

v

⋃

N

⁡

(

v

j

,

d

)

⁢

1

1

+

deg

⁡

(

v

j

)

,

where N(v j , d) represents d degree of separation from node v i .

26. The method according to claim 1 , wherein proximity prestige parameter is defined as:

PP

i

⁢

k

i



V



-

1

I

k

i

⁢

∑

j

=

1

,

j

∈

V

k

i

⁢

d

i

,

j

,

where d i,j represents a length of a shortest path between vertex v i and vertex v j , and k i represents a number of nodes in a network that can reach vertex v i .

27. The method according to claim 1 , wherein the closeness centrality parameter for the user is defined as 1/l i , where l i is defined as:

l

i

=

1



v



⁢

∑

j

∈

v

⁢

d

i

,

j

,

where d i,j represents a length of a shortest path between vertex v i and vertex v j .

28. The method according to claim 23 , wherein the discriminating analysis model accepts the determined value (v 1 ) as input, and outputs a probability value between 0 to 1, wherein, in response to determining that the probability value is greater than a predetermined threshold, the determined value (v 1 ) is determined to discriminate between churners and non-churners and used to determine the churn value (cv).

29. The method according to claim 1 , further comprising:

applying a discriminating analysis model to a plurality of graph parameters, the discriminatory analysis model outputting a probability value between 0 to 1 for each graph parameter included in the plurality of graph parameters; and

using the graph parameters from plurality of graph parameters having a probability value greater than a predetermined threshold to determine the churn value (cv).

30. The method of claim 1 , wherein

the first period of time is first one week period of time, and

the second period of time is second one week period of time that immediately follows the first one week period of time.