Systems and methods for churn prediction

Patent metadata

Field	Value
Publication number	US-10949771-B2
Application number	US-201615009603-A
Country	US
Kind code	B2
Filing date	Jan 28, 2016
Priority date	Jan 28, 2016
Publication date	Mar 16, 2021
Grant date	Mar 16, 2021

Abstract

Systems, methods, and non-transitory computer-readable media can collect past user information and churn data for a plurality of users. A churn prediction model is trained using the past user information and churn data. A churn propensity score is calculated for a present user based on the churn prediction model, the churn propensity score indicative of the likelihood of the present user to churn.

First claim

What is claimed is:

1. A computer-implemented method comprising:
collecting, by a computing system, past user information and churn data for a plurality of users, wherein the past user information comprises a plurality of user-connection entity affiliations, each user-connection entity affiliation indicative of an affiliation between a first user and a connection entity, and further wherein the churn data comprises a change of at least one of a device model, an operating system, a device brand, or a service provider;
filtering, by the computing system, the past user information based on a population-specific usage time threshold, wherein the filtering the past user information comprises:
determining a plurality of populations of users based on the past user information comprising at least one of a shared geographic area or a shared gender;
determining a population-specific usage time threshold for a first population of users of the plurality of populations of users, wherein the population-specific usage time threshold specifies a minimum usage time for a specific user-connection entity affiliation;
receiving a usage time for the first user of the first population with the specific user-connection entity affiliation; and
excluding past user information associated with the first user of the first population from the past user information for the plurality of users based on the past user information associated with the first user not satisfying the population-specific usage time threshold;

training, by the computing system, a churn prediction model with the past user information and churn data for the plurality of users; and
calculating, by the computing system, a churn propensity score for a present user for whom to predict churn propensity based on the churn prediction model, the churn propensity score indicative of the likelihood of the present user to churn.

2. The computer-implemented method of claim 1 , further comprising:
filtering, by the computing system, the past user information for the plurality of users based on a user-specific usage time threshold, wherein the filtering the past user information for the plurality of users comprises:
determining a user-specific usage time threshold for a second user; and
excluding past user information associated with the second user from the past user information for the plurality of users based on the past user information associated with the second user not satisfying the user-specific usage time threshold.

3. The computer-implemented method of claim 2 , wherein the user-specific usage time threshold specifies a percentage threshold based on a total usage time for the second user.

4. The computer-implemented method of claim 1 , wherein:
the population-specific usage time threshold specifies a minimum ranking threshold, and
the filtering the past user information further comprises:
ranking the plurality of user-connection entity affiliations based on associated usage times,
wherein the excluding the past user information further comprises excluding a second user-connection entity affiliation associated with the first user from the past user information for the plurality of users based on the second user-connection entity not satisfying the minimum ranking threshold.

5. The computer-implemented method of claim 1 , further comprising
calculating churn propensity scores for a plurality of present users for whom to predict churn propensities; and
ranking the plurality of present users based on the churn propensity scores.

6. The computer-implemented method of claim 5 , further comprising contacting a subset of the plurality of present users based on the churn propensity scores.

7. The computer-implemented method of claim 6 , wherein contacting the subset of the plurality of present users based on the churn propensity scores further comprises contacting the subset of the plurality of present users based on a churn propensity score threshold.

8. The computer-implemented method of claim 1 , wherein:
the past user information further comprises at least one of an age, a gender, an address, demographics, or a social graph.

9. A system comprising:
at least one processor; and
a memory storing instructions that, when executed by the at least one processor, cause the system to perform a method comprising:
collecting past user information and churn data for a plurality of users, wherein the past user information comprises a plurality of user-connection entity affiliations, each user-connection entity affiliation indicative of an affiliation between a first user and a connection entity, and further wherein the churn data comprises a change of at least one of a device model, an operating system, a device brand, or a service provider;

filtering the past user information based on a population-specific usage time threshold, wherein the filtering the past user information comprises:
determining a plurality of populations of users based on the past user information comprising at least one of a shared geographic area or a shared gender;
determining a population-specific usage time threshold for a first population of users of the plurality of populations of users, wherein the population-specific usage time threshold specifies a minimum usage time for a specific user-connection entity affiliation;
receiving a usage time for the first user of the first population with the specific user-connection entity affiliation; and
excluding past user information associated with the first user of the first population from the past user information for the plurality of users based on the past user information associated with the first user not satisfying the population-specific usage time threshold;

training a churn prediction model with the past user information and churn data for the plurality of users; and
calculating a churn propensity score for a present user for whom to predict churn propensity based on the churn prediction model, the churn propensity score indicative of the likelihood of the present user to churn.

10. The system of claim 9 , wherein the instructions cause the system to perform the method further comprising:
filtering the past user information for the plurality of users based on a user-specific usage time threshold, wherein the filtering the past user information for the plurality of users comprises:
determining a user-specific usage time threshold for a second user; and
excluding past user information associated with the second user from the past user information for the plurality of users based on the past user information associated with the second user not satisfying the user-specific usage time threshold.

11. The system of claim 10 , wherein the user-specific usage time threshold specifies a percentage threshold based on a total usage time for the second user.

12. The system of claim 9 , wherein:
the population-specific usage time threshold specifies a minimum ranking threshold, and
the filtering the past user information further comprises:
ranking the plurality of user-connection entity affiliations based on associated usage times,
wherein the excluding the past user information further comprises excluding a second user-connection entity affiliation associated with the first user from the past user information for the plurality of users based on the second user-connection entity not satisfying the minimum ranking threshold.

13. The system of claim 9 , wherein the instructions cause the system to perform the method further comprising:
calculating churn propensity scores for a plurality of present users for whom to predict churn propensities; and
ranking the plurality

Assignees

• Facebook Inc

Inventors

• Gonguet Vincent
• Hofleitner Aude
• Macskassy Sofus Attila
• Jarrett Steven James
• Bharathi Aruna
• Ma Zhiliang

Classifications

• G06Q10/40

  Business processes related to social networking or social networking services · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G06Q10/101

  Collaborative creation, e.g. joint development of products or services · CPC title

• G06Q50/01

  Physics · mapped topic