Disengagement prevention incentive

1 . A system, comprising:

a processor; and

a non-transitory memory storing instructions, that when executed, cause the processor to:

obtain user data and input features;

determine, by a risk identifier, predetermined risk labels based on the user data, wherein the predetermined risk labels identify disengagement risks, and determining the predetermined risk labels includes:

for each user of the user data:

determining a type of disengagement description included in the user data;

in accordance with a determination that the user data includes a first type of disengagement description, determining an explicit risk label, and

in accordance with a determination that the user data includes a second type of disengagement description, distinct from the first type of disengagement description, determining an implicit risk label;

combining each of the explicit risk labels into a set of explicit risk labels, and each of the implicit risk labels into a set of implicit risk labels; and

combining the set of explicit risk labels and the set of implicit risk labels from the user data to form the predetermined risk labels;

determine, using a first trained machine learning model that receives the input features, a first score vale characterizing a risk of disengagement for a user, wherein the first trained machine learning model is assigned to a first virtual machine that executes on a first processing resource;

determine, using a second trained machine learning model that receives the input features and the predetermined risk labels, a second score value characterizing a disengagement reason for the user, wherein the second trained machine learning model is assigned to a second virtual machine that executes on a second processing resource;

generate a risk score based on the first score value characterizing the risk of disengagement and the second score value characterizing the disengagement reason;

determine, based on the risk score, a disengagement prevention incentive for the user; and

transmit, to a computing device associated with the user, the disengagement prevention incentive, wherein the disengagement prevention incentive causes the computing device to display a user interface element for interacting with the disengagement prevention incentive.

2 . The system of claim 1 , wherein determining the explicit risk label comprises:

determining whether the first type of disengagement description includes structured data;

in accordance with a determination that the first type of disengagement description includes structured data, determining a first risk label based on the structured data;

determining whether the first type of disengagement description includes unstructured data;

in accordance with a determination that the first type of disengagement description includes unstructured data, determining a second risk label based on the unstructured data;

determining whether the first risk label and the second risk label are the same; and

in accordance with a determination that the first risk label and the second risk label are the same, including the first risk label or the second risk label in the set of explicit risk labels.

3 . The system of claim 2 , wherein determining the explicit risk label comprises:

determining whether the first risk label and the second risk label are different;

in accordance with a determination that the first risk label and the second risk label are different, determining respective scores for the first risk label and the second risk label; and

selecting a respective risk label with a score that satisfies a risk label threshold to include in the set of explicit risk labels.

4 . The system of claim 2 , wherein determining the second risk label based on the unstructured data comprises:

determining a term frequency-inverse document frequency (TF-IDF) value for the unstructured data; and

in accordance with a determination that the TF-IDF value for the unstructured data satisfies a TF-IDF threshold for a predetermined segment, associating the predetermined segment to the second risk label.

5 . The system of claim 1 , wherein determining the explicit risk label comprises:

determining whether the first type of disengagement description includes structured data; and

in accordance with a determination that the first type of disengagement description includes structured data:

determining a first risk label based on the structured data, and

including the first risk label in the set of explicit risk labels.

6 . The system of claim 1 , wherein determining the explicit risk label comprises:

determining whether the first type of disengagement description includes unstructured data; and

in accordance with a determination that the first type of disengagement description includes unstructured data:

determining a second risk label based on the unstructured data, and

including the second risk label in the set of explicit risk labels.

7 . The system of claim 1 , wherein determining the implicit risk label comprises:

determining user profile data from the second type of disengagement description;

determining an implicit risk label based on the user profile data; and

including the implicit risk label in the set of implicit risk labels.

8 . A computer-implemented method, comprising:

obtaining user data and input features;

determining, by a risk identifier, predetermined risk labels based on the user data, wherein the predetermined risk labels identify disengagement risks and determining the predetermined risk labels includes:

for each user of the user data:

determining a type of disengagement description included in the user data;

in accordance with a determination that the user data includes a first type of disengagement description, determining an explicit risk label, and

in accordance with a determination that the user data includes a second type of disengagement description, distinct from the first type of disengagement description, determining an implicit risk label;

combining each of the explicit risk labels into a set of explicit risk labels, and each of the implicit risk labels into a set of implicit risk labels; and

combining the set of explicit risk labels and the set of implicit risk labels from the user data to form the predetermined risk labels;

determining, using a first trained machine learning model that receives the input features, a first score value characterizing a risk of disengagement for a user, wherein the first trained machine learning model is assigned to a first virtual machine that executes on a first processing resource;

determining, using a second trained machine learning model that receives the input features and the predetermined risk labels, a second score value characterizing a disengagement reason for the user, wherein the second trained machine learning model is assigned to a second virtual machine that executes on a second processing resource;

generating a risk score based on the first score value characterizing the risk of disengagement and the second score value characterizing the disengagement reason;

determining, based on the risk score a disengagement prevention incentive for the user; and

transmitting, to a computing device associated with the user, the disengagement prevention incentive, wherein the disengagement prevention incentive causes the computing device to display a user interface element for interacting with the disengagement prevention incentive.

9 . The computer-implemented method of claim 8 , wherein determining the explicit risk label comprises:

determining whether the first type of disengagement description includes structured data;

in accordance with a determination that the first type of disengagement description includes structured data, determining a first risk label based on the structured data;

determining whether the first type of disengagement description includes unstructured data;

in accordance with a determination that the first type of disengagement description includes unstructured data, determining a second risk label based on the unstructured data;

determining whether the first risk label and the second risk label are the same; and

in accordance with a determination that the first risk label and the second risk label are the same, including the first risk label or the second risk label in the set of explicit risk labels.

10 . The computer-implemented method of claim 9 , wherein determining the explicit risk label comprises:

determining whether the first risk label and the second risk label are different;

in accordance with a determination that the first risk label and the second risk label are different, determining respective scores for the first risk label and the second risk label; and

selecting a respective risk label with a score that satisfies a risk label threshold to include in the set of explicit risk labels.

11 . The computer-implemented method of claim 9 , wherein determining the second risk label based on the unstructured data comprises:

determining a term frequency-inverse document frequency (TF-IDF) value for the unstructured data; and

in accordance with a determination that the TF-IDF value for the unstructured data satisfies a TF-IDF threshold for a predetermined segment, associating the predetermined segment to the second risk label.

12 . The computer-implemented method of claim 8 , wherein determining the explicit risk label comprises:

determining whether the first type of disengagement description includes structured data; and

in accordance with a determination that the first type of disengagement description includes structured data:

determining a first risk label based on the structured data, and

including the first risk label in the set of explicit risk labels.

13 . The computer-implemented method of claim 8 , wherein determining the explicit risk label comprises:

determining whether the first type of disengagement description includes unstructured data; and

in accordance with a determination that the first type of disengagement description includes unstructured data:

determining a second risk label based on the unstructured data, and

including the second risk label in the set of explicit risk labels.

14 . The computer-implemented method of claim 8 , wherein determining the implicit risk label comprises:

determining user profile data from the second type of disengagement description;

determining an implicit risk label based on the user profile data; and

including the implicit risk label in the set of implicit risk labels.

15 . A non-transitory computer readable medium having instructions stored thereon, wherein the instructions, when executed by at least one processor, cause at least one device to perform operations comprising:

obtaining user data and input features;

determining, by a risk identifier, predetermined risk labels based on the user data, wherein the predetermined risk labels identify disengagement risks and determining the predetermined risk labels includes:

for each user of the user data:

determining a type of disengagement description included in the user data;

in accordance with a determination that the user data includes a first type of disengagement description, determining an explicit risk label, and

in accordance with a determination that the user data includes a second type of disengagement description, distinct from the first type of disengagement description, determining an implicit risk label;

combining each of the explicit risk labels into a set of explicit risk labels, and each of the implicit risk labels into a set of implicit risk labels; and

combining the set of explicit risk labels and the set of implicit risk labels from the user data to form the predetermined risk labels;

determining, using a first trained machine learning model that receives the input features, a first score value characterizing a risk of disengagement for a user, wherein the first trained machine learning model is assigned to a first virtual machine that executes on a first processing resource;

determining, using a second trained machine learning model that receives the input features and the predetermined risk labels, a second score value characterizing a disengagement reason for the user, wherein the second trained machine learning model is assigned to a second virtual machine that executes on a second processing resource;

generating a risk score based on the first score value characterizing the risk of disengagement and the second score value characterizing the disengagement reason;

determining, based on the risk score a disengagement prevention incentive for the user; and

transmitting, to a computing device associated with the user, the disengagement prevention incentive, wherein the disengagement prevention incentive causes the computing device to display a user interface element for interacting with the disengagement prevention incentive.

16 . The non-transitory computer readable medium of claim 15 , wherein determining the explicit risk label comprises:

determining whether the first type of disengagement description includes structured data;

in accordance with a determination that the first type of disengagement description includes structured data, determining a first risk label based on the structured data;

determining whether the first type of disengagement description includes unstructured data;

in accordance with a determination that the first type of disengagement description includes unstructured data, determining a second risk label based on the unstructured data;

determining whether the first risk label and the second risk label are the same; and

in accordance with a determination that the first risk label and the second risk label are the same, including the first risk label or the second risk label in the set of explicit risk labels.

17 . The non-transitory computer readable medium of claim 16 , wherein determining the explicit risk label comprises:

determining whether the first risk label and the second risk label are different;

in accordance with a determination that the first risk label and the second risk label are different, determining respective scores for the first risk label and the second risk label; and

selecting a respective risk label with a score that satisfies a risk label threshold to include in the set of explicit risk labels.

18 . The non-transitory computer readable medium of claim 16 , wherein determining the second risk label based on the unstructured data comprises:

determining a term frequency-inverse document frequency (TF-IDF) value for the unstructured data; and

in accordance with a determination that the TF-IDF value for the unstructured data satisfies a TF-IDF threshold for a predetermined segment, associating the predetermined segment to the second risk label.

19 . The non-transitory computer readable medium of claim 15 , wherein determining the explicit risk label comprises:

determining whether the first type of disengagement description includes structured data; and

in accordance with a determination that the first type of disengagement description includes structured data:

determining a first risk label based on the structured data, and

including the first risk label in the set of explicit risk labels.

20 . The non-transitory computer readable medium of claim 15 , wherein determining the implicit risk label comprises:

determining user profile data from the second type of disengagement description;

determining an implicit risk label based on the user profile data; and

including the implicit risk label in the set of implicit risk labels.