Application graph builder

1. A computer-executed method for recommending applications of a predefined category to an account holder, the method comprising:

receiving information corresponding to one or more applications executing on a client device of an account holder;

generating an application graph for each account holder based on the received information, wherein the application graph is a representation of usage information of applications on the client device at least including a list of predefined application categories;

applying, for each predefined category, a computer model including a set of determined model parameters to determine a numerical likelihood that the account holder will prefer receiving applications related to the predefined category based on the generated application graph and model parameters;

wherein the determined model parameters include a set of model account holders having a positive or negative preference for the predefined category; and

recommending at least one other application of the predefined category to an other account holder, the at least one other application able to execute on an other client device of the other account holder, based on the numerical likelihood.

2. The computer-executed method of claim 1 , wherein the client device associated with a model account holder having a positive preference is an account holder that has an application of the predefined category executing on the client device.

3. The computer-executed method of claim 1 , wherein the client device associated with a model account holder having a negative preference is an account holder that has an application of the predefined category executing on the client device.

4. The computer-executed method of claim 1 , further comprising determining the set of model parameters for the computer model by:

receiving a set of metrics samples associated with previously recommended applications, the metrics samples including

a set of targeted account holders predicted to prefer applications associated with the predefined category, and

a set of accurate account holders predicted to prefer applications associated with the predefined category and having a positive preference to the application associated with the predefined category;

determining a precision function based on the comparison of the targeted account holders and accurate account holders;

updating the set of model parameters based on the precision function.

5. The computer-executed method of claim 1 , further comprising determining the set of model parameters for the computer model by:

receiving a set of metrics samples associated with previously recommended applications, the metrics samples including

a set of accurate account holders predicted to prefer applications associated with the predefined category and having a positive preference to the application associated with the predefined category, and

a set of absent account holders having a positive preference to the application associated with the predefined category;

determining a recall function based on the comparison of the accurate account holders and absent account holders;

updating the set of model parameters based on the recall function.

6. The computer-executed method of claim 1 , further comprising a requestor that sends a request for recommending applications, wherein the requestor may include an ad network, an application developer or a third-party application service.

7. The computer-executed method of claim 1 further comprising

transmitting, based on the numerical likelihood, a notification to one or more alternate applications executing on the client device of the account holder.

8. The computer-executed method of claim 1 further comprising

transmitting, based on the numerical likelihood, a notification to one or more third-party applications including advertising networks, wherein one or more third-party applications including advertising networks are sent a notification based on the likelihood.

9. The computer-executed method of claim 1 , wherein

the generated application graph is associated with a plurality of predefined categories, and

the generated application graph association is used to determine model parameters on alternate devices.

10. The computer-executed method of claim 1 , wherein

the user account of the client device is associated with a plurality of predefined categories, and

the user account association is used to determine model parameters on alternate devices.

11. The computer-executed method of claim 1 , further comprising

identifying the at least one other application to recommend based on an other application graph associated with the other account holder;

determining the identified at least one other application is able to execute on the other client device based on an the similarity between the application graph and the other application graph;

recommending the at least one other application to the other account holder based on the determination.

12. The computer-executed method of claim 1 , wherein the usage information of the client device is specific to one application executing on the client device.

13. A non-transitory computer-readable storage medium comprising instructions for recommending applications that when executed cause a processor to:

receive information corresponding to one or more applications executing on a client device of an account holder;

generate an application graph for each account holder based on the received information, wherein the application graph is a representation of usage information of applications on the client device at least including a list of predefined application categories;

apply, for each predefined category, a computer model including a set of determined model parameters to determine a numerical likelihood that the account holder will prefer receiving applications related to the predefined category based on the generated application graph and model parameters;

wherein the determined model parameters include a set of model account holders having a positive or negative preference for the predefined category; and

recommend at least one other application of the predefined category to an other account holder, the at least one other application able to execute on an other client device of the other account holder, based on the numerical likelihood.

14. The non-transitory computer readable storage medium of claim 13 , wherein the client device associated with a model account holder having a positive preference is an account holder that has an application of the predefined category executing on the client device.

15. The non-transitory computer readable storage medium of claim 13 , wherein the client device associated with a model account holder having a negative preference is an account holder that has an application of the predefined category executing on the client device.

16. The non-transitory computer readable storage medium of claim 13 , wherein the instructions further cause the processor to determine the set of model parameters for the computer model by:

receiving a set of metrics samples associated with previously recommended applications, the metrics samples including

a set of targeted account holders predicted to prefer applications associated with the predefined category, and

a set of accurate account holders predicted to prefer applications associated with the predefined category and having a positive preference to the application associated with the predefined category;

determining a precision function based on the comparison of the targeted account holders and accurate account holders;

updating the set of model parameters based on the precision function.

17. The non-transitory computer readable storage medium of claim 13 , wherein the instructions further cause the processor to determine the set of model parameters for the computer model by:

receiving a set of metrics samples associated with previously recommended applications, the metrics samples including

a set of accurate account holders predicted to prefer applications associated with the predefined category and having a positive preference to the application associated with the predefined category, and

a set of absent account holders having a positive preference to the application associated with the predefined category;

determining a recall function based on the comparison of the accurate account holders and absent account holders;

updating the set of model parameters based on the recall function.

18. The non-transitory storage medium of claim 13 , further comprising a requestor that sends a request for recommending applications, wherein the requestor may include an ad network, an application developer or a third-party application service.

19. The non-transitory computer storage medium of claim 13 , wherein the instructions further cause the processor to:

transmit, based on the numerical likelihood, a notification to one or more alternate applications executing on the client device of the account holder.

20. The non-transitory computer storage medium of claim 13 , wherein the instructions further cause the processor to:

transmit, based on the numerical likelihood, a notification to one or more third-party applications including advertising networks, wherein one or more third-party applications including advertising networks are sent a notification based on the likelihood.

21. The non-transitory computer storage medium of claim 13 , wherein

the generated application graph is associated with a plurality of predefined categories, and

the generated application graph association is used to determine model parameters on alternate devices.

22. The non-transitory computer storage medium of claim 13 , wherein

the user account of the client device is associated with a plurality of predefined categories, and

the user account association is used to determine model parameters on alternate devices.

23. The non-transitory computer storage medium of claim 13 , wherein the instructions further cause the processor to:

identify the at least one other application to recommend based on an other application graph associated with the other account holder;

determine the identified at least one other application is able to execute on the other client device based on an the similarity between the application graph and the other application graph;

recommend the at least one other application to the other account holder based on the determination.

24. The non-transitory computer storage medium of claim 13 , wherein the usage information of the client device is specific to one application executing on the client device.

25. A system comprising a processor and a memory storing computer program instructions for recommending applications that when executed by the processor cause the processor to:

receive information corresponding to one or more applications executing on a client device of an account holder;

generate an application graph for each account holder based on the received information, wherein the application graph is a representation of usage information of applications on the client device at least including a list of predefined application categories;

apply, for each predefined category, a computer model including a set of determined model parameters to determine a numerical likelihood that the account holder will prefer receiving applications related to the predefined category based on the generated application graph and model parameters;

wherein the determined model parameters include a set of model account holders having a positive or negative preference for the predefined category; and

recommend at least one other application of the predefined category to an other account holder, the at least one other application able to execute on an other client device of the other account holder, based on the numerical likelihood.

26. The system of claim 25 , wherein the client device associated with a model account holder having a positive preference is an account holder that has an application of the predefined category executing on the client device.

27. The system of claim 25 , wherein the client device associated with a model account holder having a negative preference is an account holder that has an application of the predefined category executing on the client device.

28. The system of claim 25 , wherein the instructions further cause the processor to determine the set of model parameters for the computer model by:

receiving a set of metrics samples associated with previously recommended applications, the metrics samples including

a set of targeted account holders predicted to prefer applications associated with the predefined category, and

a set of accurate account holders predicted to prefer applications associated with the predefined category and having a positive preference to the application associated with the predefined category;

determining a precision function based on the comparison of the targeted account holders and accurate account holders;

updating the set of model parameters based on the precision function.

29. The system of claim 25 , wherein the instructions further cause the processor to determine the set of model parameters for the computer model by:

receiving a set of metrics samples associated with previously recommended applications, the metrics samples including

a set of accurate account holders predicted to prefer applications associated with the predefined category and having a positive preference to the application associated with the predefined category, and

a set of absent account holders having a positive preference to the application associated with the predefined category;

determining a recall function based on the comparison of the accurate account holders and absent account holders;

updating the set of model parameters based on the recall function.

30. The system of claim 25 , further comprising a requestor that sends a request for recommending applications, wherein the requestor may include an ad network, an application developer or a third-party application service.

31. The system of claim 25 , wherein the instructions further cause the processor to:

transmit, based on the numerical likelihood, a notification to one or more alternate applications executing on the client device of the account holder.

32. The system of claim 25 , wherein the instructions further cause the processor to:

transmit, based on the numerical likelihood, a notification to one or more third-party applications including advertising networks, wherein one or more third-party applications including advertising networks are sent a notification based on the likelihood.

33. The system of claim 25 , wherein

the generated application graph is associated with a plurality of predefined categories, and

the generated application graph association is used to determine model parameters on alternate devices.

34. The system of claim 25 , wherein

the user account of the client device is associated with a plurality of predefined categories, and

the user account association is used to determine model parameters on alternate devices.

35. The system of claim 25 wherein the instructions further cause the processor to:

identify the at least one other application to recommend based on an other application graph associated with the other account holder;

determine the identified at least one other application is able to execute on the other client device based on an the similarity between the application graph and the other application graph;

recommend the at least one other application to the other account holder based on the determination.

36. The system of claim 25 , wherein the usage information of the client device is specific to one application executing on the client device.