Automatic recipient targeting for notifications

1. A method comprising, by one or more computing devices associated with a social-networking system:

detecting a triggering action by a user of the social-networking system, wherein the detecting comprises receiving information about the triggering action from a client device associated with the user, wherein the triggering action comprises:

a click-through of an icon by the user,

a mouseover of an icon by the user, or

a request to log in to the social-networking system by the user;

accessing, in response to detection of the triggering action, a queue comprising a plurality of notifications, wherein each of the notifications was received by the social-networking system from an application associated with the social-networking system prior to the detecting the triggering action, and wherein the detected triggering action is associated with one or more of the notifications in the queue;

for each of the one or more notifications in the queue associated with the triggering action:

calculating, using a machine-learning model, a click-through probability that the user will interact with the notification upon display of the notification, wherein the machine-learning model is based at least in part on one or more features associated with the user or the notification, and wherein the features comprise a period between receipt of the notification and the detecting of the triggering action;

determining whether the click-through probability satisfies a threshold; and

if the click-through probability satisfies the threshold, then sending the notification to the client device associated with the user for display;

else, removing the notification from the queue.

2. The method of claim 1 , wherein the features further comprise, for an application associated with the notification:

the user's frequency of interacting with notifications associated with the social-networking system within a specified timeframe;

the user's frequency of interacting with the application within a specified timeframe;

the user's click-through rate for notifications associated with the social-networking system;

the user's click-through rate for notifications associated with the application;

a time of the user's most recent interaction with the application; or

a category corresponding to the application.

3. The method of claim 1 , wherein the features comprise:

the user's frequency of interacting with applications associated with the social-networking system, wherein the frequency is assessed within a specified timeframe;

the user's frequency of interacting with applications associated with the social-networking system, wherein the applications fall within a particular category;

a time of the user's most recent interaction with any application associated with the social-networking system; or

a time of the user's most recent interaction with an application associated with the social-networking system, wherein the application falls within a particular category.

4. The method of claim 1 , wherein the features comprise, for an application associated with the notification:

a number of daily active users of the application;

a number of monthly active users of the application;

a growth rate of the application's number of daily active users;

a growth rate of the application's number of monthly active users;

a conversion rate associated with the application among users of the social-networking system fitting a particular demographic description; or

a language supported by the application.

5. The method of claim 1 , further comprising:

computing one or more similarity scores between an application associated with at least one of the accessed notifications and one or more other applications;

identifying one or more similar applications based on the computed similarity scores; and

extracting one or more features based on the user's interactions with the identified similar applications.

6. The method of claim 1 , wherein the determining whether the click-through probability satisfies a threshold further comprises, for at least one of the accessed notifications:

determining whether the triggering action took place within a timeframe within which the notification is valid.

7. The method of claim 1 , further comprising:

detecting one or more actions taken by the user upon display of one or more of the accessed notifications; and

modifying the machine-learning model based on the detected actions.

8. One or more computer-readable non-transitory storage media embodying software that is operable when executed to:

detect a triggering action by a user of the social-networking system, wherein the detecting comprises receiving information about the triggering action from a client device associated with the user, wherein the triggering action comprises:

a click-through of an icon by the user,

a mouseover of an icon by the user, or

a request to log in to the social-networking system by the user;

access, in response to detection of the triggering action, a queue comprising a plurality of notifications, wherein each of the notifications was received by the social-networking system from an application associated with the social-networking system prior to the detecting the triggering action, and wherein the detected triggering action is associated with one or more of the notifications in the queue;

for each of the one or more notifications in the queue associated with the triggering action:

calculate, using a machine-learning model, a click-through probability that the user will interact with the notification upon display of the notification, wherein the machine-learning model is based at least in part on one or more features associated with the user or the notification, and wherein the features comprise a period between receipt of the notification and the detecting of the triggering action;

determine whether the click-through probability satisfies a threshold; and

if the click-through probability satisfies the threshold, then send the notification to the client device associated with the user for display;

else, remove the notification from the queue.

9. The media of claim 8 , wherein the features comprise, for an application associated with the notification:

the user's frequency of interacting with notifications associated with the social-networking system within a specified timeframe;

the user's frequency of interacting with the application within a specified timeframe;

the user's click-through rate for notifications associated with the social-networking system;

the user's click-through rate for notifications associated with the application;

a time of the user's most recent interaction with the application; or

a category corresponding to the application.

10. The media of claim 8 , wherein the features comprise:

the user's frequency of interacting with applications associated with the social-networking system, wherein the frequency is assessed within a specified timeframe;

the user's frequency of interacting with applications associated with the social-networking system, wherein the applications fall within a particular category;

a time of the user's most recent interaction with any application associated with the social-networking system; or

a time of the user's most recent interaction with an application associated with the social-networking system, wherein the application falls within a particular category.

11. The media of claim 8 , wherein the features comprise, for an application associated with the notification:

a number of daily active users of the application;

a number of monthly active users of the application;

a growth rate of the application's number of daily active users;

a growth rate of the application's number of monthly active users;

a conversion rate associated with the application among users of the social-networking system fitting a particular demographic description; or

a language supported by the application.

12. The media of claim 8 , wherein the software is further operable when executed to:

compute one or more similarity scores between an application associated with at least one of the accessed notifications and one or more other applications;

identify one or more similar applications based on the computed similarity scores; and

extract one or more features based on the user's interactions with the identified similar applications.

13. The media of claim 8 , wherein the determining whether the click-through probability satisfies a threshold further comprises, for at least one of the accessed notifications:

determining whether the triggering action took place within a timeframe within which the notification is valid.

14. The media of claim 8 , wherein the software is further operable when executed to:

detect one or more actions taken by the user upon display of one or more of the accessed notifications; and

modify the machine-learning model based on the detected actions.

15. A system comprising: one or more processors; and a memory coupled to the processors comprising instructions executable by the processors, the processors being operable when executing the instructions to:

detect a triggering action by a user of the social-networking system, wherein the detecting comprises receiving information about the triggering action from a client device associated with the user, wherein the triggering action comprises:

a click-through of an icon by the user,

a mouseover of an icon by the user, or

a request to log in to the social-networking system by the user;

access, in response to detection of the triggering action, a queue comprising a plurality of notifications, wherein each of the notifications was received by the social-networking system from an application associated with the social-networking system prior to the detecting the triggering action, and wherein the detected triggering action is associated with one or more of the notifications in the queue;

for each of the one or more notifications in the queue associated with the triggering action:

calculate, using a machine-learning model, a click-through probability that the user will interact with the notification upon display of the notification, wherein the machine-learning model is based at least in part on one or more features associated with the user or the notification, and wherein the features comprise a period between receipt of the notification and the detecting of the triggering action;

determine whether the click-through probability satisfies a threshold; and

if the click-through probability satisfies the threshold, then send the notification to the client device associated with the user for display;

else, remove the notification from the queue.

16. The system of claim 15 , wherein the features comprise, for an application associated with the notification:

the user's frequency of interacting with notifications associated with the social-networking system within a specified timeframe;

the user's frequency of interacting with the application within a specified timeframe;

the user's click-through rate for notifications associated with the social-networking system;

the user's click-through rate for notifications associated with the application;

a time of the user's most recent interaction with the application; or

a category corresponding to the application.

17. The system of claim 15 , wherein the features comprise:

the user's frequency of interacting with applications associated with the social-networking system, wherein the frequency is assessed within a specified timeframe;

the user's frequency of interacting with applications associated with the social-networking system, wherein the applications fall within a particular category;

a time of the user's most recent interaction with any application associated with the social-networking system; or

a time of the user's most recent interaction with an application associated with the social-networking system, wherein the application falls within a particular category.