System and method for generating subscriber churn predictions

1. A system for generating a subscriber churn prediction comprising:

a data input device adapted to receive call detail records from a network operator detailing communication for at least one subscriber of the network operator; and

at least one processor connected to the data input device, the at least one processor adapted to execute a churn prediction program that determines tie-strengths between connected subscribers of the network operator based on a weighted sum of a plurality of normalized call attributes from the call detail records and derives at least one social metric with the tie-strengths;

wherein the churn prediction program generates the subscriber churn prediction based at least on the at least one social metric.

2. The system according to claim 1 , wherein the churn prediction program includes a social network analysis module that determines the tie-strengths and derives social metrics from the call detail records for generating the churn prediction.

3. The system according to claim 2 , wherein the churn prediction program includes a churn prediction module that receives the social metrics from the social network analysis module and generates the churn prediction.

4. The system according to claim 3 , wherein the churn prediction is generated from the social metrics in combination with at least one of a service usage metric, a billing metric, a number of calls to customer support, an outcome of complaints or a demographic data.

5. The system according to claim 1 , wherein the at least one social metric includes at least one of a net churner influence, a number of neighboring subscribers that are churners, a number of hops to a nearest churner, a number of calls to churners, a number of calls to the nearest churner, and a time spent on calls to churners.

6. The system according to claim 1 , wherein the social network analysis module propagates a churner influence between subscribers based on the tie-strength and determines a net influence for each subscriber.

7. The system according to claim 6 , wherein the propagation of the churner influence is receiver centric with the influence received by each subscriber from a connected subscriber being proportional to the tie-strength of the tie between the subscribers.

8. A computerized method for generating a prediction of subscriber churn comprising the steps of:

receiving, by a churn prediction program executing on a computer processor, call detail records from a network operator detailing communication for at least one subscriber of the network operator;

determining, by the churn prediction program executing on the computer processor, tie-strengths between connected subscribers on the network operator based on a weighted sum of a plurality of normalized call attributes from the call detail records;

deriving, by the churn prediction program executing on the computer processor, at least one social metric with the tie-strengths; and

generating, by the churn prediction program executing on the computer processor, a subscriber churn prediction based at least on the at least one social metric.

9. The computerized method according to claim 8 , wherein the step of generating, by the churn prediction program executing on the computer processor, the subscriber churn prediction includes combining the social metrics with at least one of a service usage metric, a billing metric, a number of calls to customer support, an outcome of complaints or a demographic data.

10. The computerized method according to claim 8 , wherein the at least one social metric includes at least one of a net churner influence, a number of neighboring subscribers that are churners, a number of hops to a nearest churner, a number of calls to churners, a number of calls to the nearest churner, and a time spent on calls to churners.

11. The computerized method according to claim 8 , additionally comprising the step of:

propagating, by the churn prediction program executing on the computer processor, a churner influence between subscribers based on the tie-strength and determining a net influence for each subscriber.

12. The computerized method according to claim 11 , wherein propagating, by the churn prediction program executing on the computer processor, the churner influence between subscribers is receiver centric with the influence received by each subscriber from a connected subscriber being proportional to the tie-strength of the tie between the subscribers.

13. A non-transitory, tangible computer-readable medium storing instructions adapted to be executed by a computer processor to perform a method for generating a subscriber churn prediction, said method comprising the steps of:

receiving, by a churn prediction program executing on the computer processor, call detail records from a network operator detailing communication for at least one subscriber of the network operator;

determining, by the churn prediction program executing on the computer processor, tie-strengths between connected subscribers on the network operator based on a weighted sum of a plurality of normalized call attributes from the call detail records;

deriving, by the churn prediction program executing on the computer processor, at least one social metric with the tie-strengths; and

generating, by the churn prediction program executing on the computer processor, a subscriber churn prediction based at least on the at least one social metric.

14. The non-transitory, tangible computer-readable medium of claim 13 , wherein the step of generating, by the churn prediction program executing on the computer processor, the subscriber churn prediction includes combining the social metrics with at least one of a service usage metric, a billing metric, a number of calls to customer support, an outcome of complaints or a demographic data.

15. The non-transitory, tangible computer-readable medium of claim 13 , wherein the method further comprises the step of:

propagating, by the churn prediction program executing on the computer processor, a churner influence between subscribers based on the tie-strength and determining a net influence for each subscriber.