Using machine learning to customize notifications for users
A method includes extracting data pertaining to a plurality of user actions in connection with one or more changes to one or more of a plurality of applications, and training one or more machine learning models with the extracted data. The one more machine learning models are used to predict whether a user should receive a given notification in connection with a given change to a given application of the plurality of applications. In response to predicting that the user should receive the given notification, content of the given notification is determined. The method further includes generating the given notification for the user, and transmitting the given notification to the user.
1. A method, comprising:
extracting data pertaining to a plurality of user actions in connection with one or more changes to one or more of a plurality of applications;
training one or more machine learning models with the extracted data;
using the one more machine learning models to predict whether a user should receive a given notification in connection with a given change to a given application of the plurality of applications;
in response to predicting that the user should receive the given notification, determining, using the one or more machine learning models, content of the given notification;
generating the given notification for the user; and
transmitting the given notification to the user;
wherein the determining of the content of the given notification comprises:
identifying one or more notifications received by the user as being at least similar to the given notification;
determining that the one or more notifications received by the user resulted in one or more support requests; and
including one or more details in the given notification to address one or more issues with the given change to the given application responsive to the determination that the one or more notifications received by the user resulted in the one or more support requests;
wherein extracting the data comprises calculating a ratio of a number of times the user contacted technical support to a number associated with a read status of the user with respect to a corresponding notification in connection with respective ones of the one or more changes;
wherein the number associated with the read status of the user comprises one of a number of times the user failed to read and a number of times the user read a corresponding notification in connection with respective ones of the one or more changes; and
wherein the steps of the method are executed by a processing device operatively coupled to a memory.
2. The method of claim 1 , wherein the one or more changes to the one or more of the plurality of applications comprise at least one of an installation of and an upgrade to the one or more of the plurality of applications.
3. The method of claim 1 , wherein extracting the data further comprises collecting one or more clickstreams of the user.
4. The method of claim 1 , wherein extracting the data further comprises calculating a ratio of a number of notifications read by the user to a number of notifications received by the user in connection with respective ones of the one or more changes.
5. The method of claim 2 , further comprising:
extracting data about the one or more changes to the one or more of a plurality of applications; and
training the one or more machine learning models with the extracted data about the one or more changes.
6. The method of claim 5 , wherein extracting the data about the one or more changes comprises calculating at least one of an overall average installation time and an overall average upgrade time for the plurality of applications.
7. The method of claim 5 , wherein extracting the data about the one or more changes comprises calculating at least one of an average installation time and an average upgrade time for one or more subsets of the plurality of applications determined to have the same application type.
8. The method of claim 5 , wherein extracting the data about the one or more changes comprises calculating at least one of: (i) a ratio of a number of successful application upgrades to a total number of attempted application upgrades; and (ii) a ratio of a number of successful application installations to a total number of attempted application installations.
9. The method of claim 1 , further comprising:
determining one or more characteristics of respective ones of the plurality of applications; and
identifying, using cosine similarity, two or more of the plurality of applications that are similar to each other based on the determined one or more characteristics.
10. The method of claim 9 , wherein the prediction whether the user should receive the given notification in connection with the given change to the given application is based at least in part on data pertaining to one or more user actions in connection with one or more of the plurality of applications identified as similar to the given application.
11. The method of claim 1 , wherein the one or more machine learning models comprise a plurality of decision trees, and the plurality of decision trees are respectively trained with different portions of the extracted data.
12. The method according to claim 11 , wherein:
each of the plurality of decision trees yields one of a positive result and a negative result with respect to whether the user should receive the given notification; and
the prediction whether the user should receive the given notification corresponds to the result produced by a majority of the plurality of decision trees.
13. An apparatus comprising: a processing device operatively coupled to a memory and configured to: extract data pertaining to a plurality of user actions in connection with one or more changes to one or more of a plurality of applications; train one or more machine learning models with the extracted data; use the one more machine learning models to predict whether a user should receive a given notification in connection with a given change to a given application of the plurality of applications; in response to predicting that the user should receive the given notification, determine, using the one or more machine learning models, content of the given notification; generate the given notification for the user; and transmit the given notification to the user; wherein, in determining the content of the given notification, the processing device is configured to: identify one or more notifications received by the user as being at least similar to the given notification; determine that the one or more notifications received by the user resulted in one or more support requests; and include one or more details in the given notification to address one or more issues with the given change to the given application responsive to the determination that the one or more notifications received by the user resulted in the one or more support requests; wherein, in extracting the data, the processing device is configured to calculate a ratio of a number of times the user contacted technical support to a number associated with a read status of the user with respect to a corresponding notification in connection with respective ones of the one or more changes; and wherein the number associated with the read status of the user comprises one of a number of times the user failed to read and a number of times the user read a corresponding notification in connection with respective ones of the one or more changes.
14. The apparatus of claim 13 , wherein the one or more changes to the one or more of the plurality of applications comprise at least one of an installation of and an upgrade to the one or more of the plurality of applications.
15. The apparatus of claim 13 , wherein the processing device is further configured to:
determine one or more characteristics of respective ones of the plurality of applications; and
identify, using cosine similarity, two or more of the plurality of applications that are similar to each other based on the determined one or more characteristics.
16. The apparatus of claim 15 , wherein the prediction whether the user should receive the given notification in connection with the given change to the given application is based at least in part on data pertaining to one or more user actions in connection with one or more of the plurality of applications identified as similar to the given application.
17. An article of manufacture comprising a non-transitory processor-readable storage medium having stored therein program code of one or more software programs, wherein the program code when executed by at least one processing device causes said at least one processing device to perform the steps of: extracting data pertaining to a plurality of user actions in connection with one or more changes to one or more of a plurality of applications; training one or more machine learning models with the extracted data; using the one more machine learning models to predict whether a user should receive a given notification in connection with a given change to a given application of the plurality of applications; in response to predicting that the user should receive the given notification, determining, using the one or more machine learning models, content of the given notification; generating the given notification for the user; and transmitting the given notification to the user; wherein, in determining the content of the given notification, the program code causes said at least one processing device to perform the steps of: identifying one or more notifications received by the user as being at least similar to the given notification; determining that the one or more notifications received by the user resulted in one or more support requests; and including one or more details in the given notification to address one or more issues with the given change to the given application responsive to the determination that the one or more notifications received by the user resulted in the one or more support requests; wherein, in extracting the data, the program code causes said at least one processing device to calculate a ratio of a number of times the user contacted technical support to a number associated with a read status of the user with respect to a corresponding notification in connection with respective ones of the one or more changes; and wherein the number associated with the read status of the user comprises one of a number of times the user failed to read and a number of times the user read a corresponding notification in connection with respective ones of the one or more changes.
18. The article of manufacture of claim 17 , the one or more changes to the one or more of the plurality of applications comprise at least one of an installation of and an upgrade to the one or more of the plurality of applications.
19. The article of manufacture of claim 17 , wherein the program code further causes said at least one processing device to perform the steps of: determining one or more characteristics of respective ones of the plurality of applications; and identifying, using cosine similarity, two or more of the plurality of applications that are similar to each other based on the determined one or more characteristics.
20. The article of manufacture of claim 19 , wherein the prediction whether the user should receive the given notification in connection with the given change to the given application is based at least in part on data pertaining to one or more user actions in connection with one or more of the plurality of applications identified as similar to the given application.