Proactive customer care in a communication system

1. A device, comprising:

a processing system including a processor; and

a memory that stores executable instructions that, when executed by the processing system, facilitate performance of operations, the operations comprising:

retrieving historical voice call information and historical data call information for a plurality of customers, including a selected customer, of a communication service provider, wherein the historical voice call information and historical data call information comprises information about network access by the plurality of customers to a communication network of the communication service provider;

training a machine learning model using training data based on the historical voice call information and the historical data call information,

wherein the training the machine learning model comprises defining, using a sliding window, a plurality of slices, each respective slice of the plurality of slices segmented to include a plurality of time units, each respective time unit including associated historical voice call information and associated historical data call information,

wherein at least some respective time units of the respective slices of the plurality of slices overlap in time to reuse the associated historical voice call information and associated historical data call information to form the training data for the machine learning model;

applying current voice call information for the selected customer and current data call information for the selected customer to the machine learning model to determine a likelihood that the selected customer will require customer care assistance by the communication service provider; and

responsive to the likelihood that the selected customer will require customer care assistance exceeding a predetermined threshold, initiating a corrective action on the communication network of the communication service provider to reduce the likelihood that the selected customer will require customer care assistance.

2. The device of claim 1 , wherein the operations further comprise:

identifying a customer device used by the selected customer to access the communication network of the communication service provider; and

initiating, by the communication network, a cancelation of a registration of the customer device with the communication network as a correction action.

3. The device of claim 1 , wherein the operations further comprise:

applying a sliding window feature extraction process to the historical voice call information and the historical data call information to remove seasonality effects in the historical voice call information and the historical data call information.

4. The device of claim 3 , wherein the operations further comprise:

applying a second sliding window feature extraction process to the current voice call information and the current data call information before applying the current voice call information and the current data call information to the machine learning model, wherein the second sliding window feature extraction process differs from the sliding window feature extraction process.

5. The device of claim 1 , wherein retrieving the historical voice call information and the historical data call information comprises retrieving a care log dataset and retrieving a network log dataset, wherein retrieving a care log dataset comprises receiving information about a plurality of customer care calls including information about a customer associated with a respective customer care call, information about a customer care issue associated with the respective customer care call and information about actions taken for resolving the customer care issue associated with the respective customer care call, and wherein retrieving a network log dataset comprises receiving information about a plurality of data communication sessions including information about a customer associated with a respective data communication session and information about a cause of termination of the respective data communication session.

6. The device of claim 1 , wherein the operations further comprise applying the current voice call information for the selected customer and current data call information for the selected customer to an aggregated features model (AFM) to determine a first likelihood that the selected customer will require customer care assistance.

7. The device of claim 6 , wherein the operations further comprise applying the current voice call information for the selected customer and current data call information for the selected customer to an individual variations model (IVM) to determine a second likelihood that the selected customer will require customer care assistance.

8. The device of claim 7 , wherein the operations further comprise:

applying the first likelihood that the selected customer will require customer care assistance and the second likelihood that the selected customer will require customer care assistance to a feedback model (FBM); and

receiving from the feedback model a value corresponding to the likelihood that the selected customer will require customer care assistance.

9. The device of claim 1 , wherein the device comprises one device of a plurality of similar devices, each respective device of the plurality of similar devices is dedicated to a respective unique portion of the communication network of the communication service provider.

10. The device of claim 1 , wherein the operations further comprise:

extracting feature vectors from the historical voice call information and historical data call information.

11. A non-transitory computer-readable medium, comprising executable instructions that, when executed by a processing system including a processor, facilitate performance of operations, the operations comprising:

receiving care log information and network log information about communications between a communication network of a communication service provider and a plurality of customers of the communication service provider;

extracting feature vectors from the care log information and the network log information including defining, using a sliding window, a plurality of slices, each respective slice of the plurality of slices segmented to include a plurality of time units, each respective time unit including associated care log information and associated network log information,

wherein at least some respective time units of the respective slices of the plurality of slices overlap in time to reuse the associated care log information and the associated network log information to form the feature vectors;

training a machine learning model using the feature vectors;

providing current call information to the machine learning model, wherein the providing current call information comprises providing information about current calls for a selected customer of the plurality of customers, wherein the information about current calls comprises current voice call information for the selected customer and current data call information for the selected customer for an analysis window;

receiving, from the machine learning model, information about a likelihood that the selected customer will require customer care assistance by the communication service provider; and

responsive to the likelihood that the selected customer will require customer care assistance exceeding a predetermined threshold, initiating a corrective action on the communication network of the communication service provider to reduce the likelihood that the selected customer will require customer care assistance.

12. The non-transitory, computer-readable medium of claim 11 , wherein the operations further comprise:

receiving, from the machine learning model, information about a likelihood that a plurality of customers, including the selected customer, will require customer care assistance; and

based on the likelihood that the plurality of customers will require customer care assistance, providing a list of telephone numbers, including a telephone number of the selected customer.

13. The non-transitory, computer-readable medium of claim 12 , wherein the operations further comprise . . .

based on the telephone number of the selected customer, identifying a customer device used by the selected customer to access the communication network of the communication service provider; and

initiating, by the communication network, a registration cancelation of the customer device and the communication network as a correction action.

14. The non-transitory, computer-readable medium of claim 11 , wherein extracting feature vectors from the care log information and the network log information comprises:

performing a sliding window feature extraction operation on the care log information and the network log information to develop a prediction for a designated time period.

15. The non-transitory, computer-readable medium of claim 14 , wherein the operations further comprise:

providing the current call information for the selected customer to a second sliding window feature extraction operation.

16. The non-transitory, computer-readable medium of claim 11 , wherein the operations further comprise:

providing the current call information for the selected customer of the plurality of customers a first model to generate a first probability, wherein the first model aggregates features of the current call information;

providing the current call information for the selected customer of the plurality of customers to a second model to generate a second probability, wherein the second model averages usage of the communication network of the communication service provider by the selected customer; and

combining the first probability and the second probability to produce the information about a likelihood that the selected customer will require customer care assistance.

17. A method, comprising:

automatically detecting, by a processing system including a processor, a service issue of a telecommunication system that may impact a customer of a telecommunication service provider of the telecommunication system;

training, by the processing system, a machine learning model using training data, the training data formed by defining, using a sliding window, a plurality of slices, each respective slice of the plurality of slices segmented to include a plurality of time units, each respective time unit including associated historical voice call information and associated historical data call information of the telecommunication system,

wherein at least some respective time units of the respective slices of the plurality of slices overlap in time to reuse the associated historical voice call information and associated historical data call information to form the training data for training the machine learning model

predicting, by the processing system implementing the machine learning model, a future customer care interaction by the customer as a result of the service issue;

initiating, by the processing system, a resolution action of the telecommunication system for the service issue before the customer contacts a customer care agent of the telecommunication service provider, wherein the initiating is responsive to the predicting the future customer care interaction; and

modifying a component of the telecommunication system to improve efficiency of operation of the telecommunication system by reducing a number of customer care contacts by customers of the telecommunication service provider.

18. The method of claim 17 , further comprising:

retrieving, by the processing system, historical data about operation of the telecommunication system, the historical data comprising the associated historical voice call information and the associated historical data call information;

extracting, by the processing system, feature data from the historical data to generate feature vectors; and

training, by the processing system, the machine learning model using the feature vectors.

19. The method of claim 18 , wherein the predicting a further customer care interaction comprises:

determining, by the processing system, a likelihood that the customer will initiate the future customer care interaction using the machine learning model.

20. The method of claim 18 , wherein the retrieving historical data about operation of the telecommunication system comprises:

receiving, by the processing system, information about a plurality of customer care calls including information about a customer associated with a respective customer care call, information about a customer care issue associated with the respective customer care call and information about actions taken for resolving the customer care issue associated with the respective customer care call; and

receiving, by the processing system, information about a plurality of data communication sessions including information about a customer associated with a respective data communication session and information about a cause of termination of the respective data communication session.