AUTOMATIC CUSTOMER COMPLAINT RESOLUTION

• US 20170019315A1
• Filed: 07/15/2016
• Published: 01/19/2017
• Est. Priority Date: 07/15/2015
• Status: Active Grant

• Current Assignee: Tupl, Inc.
• Original Assignee: Tupl, Inc.
• Inventors: Tapia, Pablo, Sanchez-Mejias, Rafael
• Granted Patent: US 10,063,406 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: H04L 41/0631 using root cause analysis; using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis
• CPC Class Codes: H04L 41/16 using machine learning or artificial intelligence
• CPC Class Codes: H04L 41/5067 Customer-centric QoS measurements
• CPC Class Codes: H04L 41/5074 Handling of user complaints or trouble tickets
• CPC Class Codes: H04W 24/08 Testing, supervising or monitoring using real traffic
• CPC Class Codes: H04W 24/10 Scheduling measurement reports ; Arrangements for measurement reports

Abstract

An analytic application may automatically determine a root cause of an issue with a wireless carrier network and generate a solution for the root cause. Initially, a data management platform may receive performance data regarding user device and network components of a wireless carrier network from multiple data sources. Subsequently, the analytic application may receive an indication of an issue affecting one or more user devices that are using the wireless carrier network. The analytic application may analyze the performance data using a trained machine learning model to determine a root cause for the issue affecting the one or more user devices. The trained machine learning model may employ multiple types of machine learning algorithms to analyze the performance data. The analytic application may provide the root cause or the solution that resolves the root cause for presentation.

Claims

1. One or more non-transitory computer-readable media storing computer-executable instructions that upon execution cause one or more processors to perform acts comprising:

receiving performance data regarding user device and network components of a wireless carrier network from multiple data sources;

receiving an indication of an issue affecting one or more user devices that are using the wireless carrier network, the issue impacting a quality of service experienced by at least one user of the one or more user devices as the one or more user devices are used to make voice calls, make multimedia calls, upload data, or download data using the wireless carrier network;

analyzing the performance data using a trained machine learning model to determine a root cause for the issue affecting the one or more user devices, the trained machine learning model employing multiple types of machine learning algorithms to analyze the performance data; and

providing at least one of the root cause and a solution that resolves the root cause for presentation.

2. The one or more non-transitory computer-readable media of claim 1, wherein the acts further comprise:

analyzing the performance data using the trained machine learning model to predict a potential issue for one or more additional user devices;

notifying at least one additional subscriber of the potential issue with the one or more additional user devices; and

tracking performance of the one or more additional user devices to detect occurrence of the potential issue on at least one additional user device.

3. The one or more non-transitory computer-readable media of claim 1, wherein the performance data includes one or more of network component performance data, user device performance data, social media data, alarm data, trouble ticket data, or key performance indicator data generated by a network monitoring tool.

4. The one or more non-transitory computer-readable media of claim 1, wherein the performance data includes multiple sets of performance data that are aggregated according to one or more grouping parameters or converged according to a unitary storage schema.

5. The one or more non-transitory computer-readable media of claim 4, wherein the one or more grouping parameters include a specific time period, a specific network component, a specific user device vendor, a specific user device model, or different levels of a network hierarchy that includes a subscriber level, a device level, a service area level, and a geographical market level, wherein the geographical market level further includes a zip code sublevel, a municipality sublevel, or an additional location-based level.

6. The one or more non-transitory computer-readable media of claim 1, wherein the receiving an indication includes receiving a subscriber or a network monitoring application generated trouble ticket.

7. The one or more non-transitory computer-readable media of claim 1, wherein the analyzing includes analyzing real time performance data or non-real time performance data for a particular time interval.

8. The one or more non-transitory computer-readable media of claim 1, wherein the analyzing includes tracking geolocations of a user as the user roams between the geolocations with a user device during a particular time interval, and analyzing the performance data that are related to the geolocations to determine the root cause.

9. The one or more non-transitory computer-readable media of claim 1, wherein the analyzing includes analyzing performance of various network cells to generate a list of one or more top offender cells that contributes to the issue.

10. The one or more non-transitory computer-readable media of claim 1, wherein the providing includes providing a particular solution that is retrieved from a solutions database for the root cause.

11. The one or more non-transitory computer-readable media of claim 1, wherein the providing includes providing a visualization of at least one of the root cause or the solution with reference to one or more specific subscribers, one or more specific user devices, one or more specific geolocations, one or more specific geographical areas, or one or more specific network cells.

12. A computer-implemented method, comprising:

receiving, at a data management platform executing on one or more computing nodes, performance data regarding user device and network components of a wireless carrier network from multiple data sources, the performance data includes one or more of network component performance data, user device performance data, social media data, alarm data, trouble ticket data, or key performance indicator data generated by a network monitoring tool;

receiving, at an analytic application executing on the one or more computing nodes, an indication of an issue affecting one or more user devices that are using the wireless carrier network, the issue impacting a quality of service experienced by at least one user of the one or more user devices as the one or more user devices are used to make voice calls, make multimedia calls, upload data, or download data using the wireless carrier network, the indication being a subscriber or a network monitoring application generated trouble ticket;

analyzing, via the analytic application executing on the one or more computing nodes, the performance data using a trained machine learning model to determine a root cause for the issue affecting the one or more user devices, the trained machine learning model employing multiple types of machine learning algorithms to analyze the performance data; and

providing, via the analytic application executing on the one or more computing nodes, at least one of the root cause and a solution that resolves the root cause for presentation.

13. The computer-implemented method of claim 12, wherein the analyzing includes tracking geolocations of a user as the user roams between the geolocations with a user device during a particular time interval, and analyzing the performance data that are related to the geolocations to determine the root cause.

14. The computer-implemented method of claim 12, wherein the analyzing includes analyzing performance of various network cells to generate a list of one or more top offender cells that contributes to the issue.

15. The computer-implemented method of claim 12, wherein the providing includes providing a visualization of at least one of the root cause or the solution with reference to one or more specific subscribers, one or more specific user devices, one or more specific geolocations, one or more specific geographical areas, or one or more specific network cells.

16. The computer-implemented method of claim 12, further comprising:

performing feature engineering on a training corpus for generating a machine learning model that is used to determine solutions for issues with the wireless carrier network;

training an initial type of machine learning algorithm using the training corpus to generate the machine learning model;

determining a training error measurement of the machine learning model is above a training error threshold;

selecting an additional type of machine learning algorithm based on a magnitude of the training error measurement according to one or more algorithm selection rules in response to the training error measurement being above the training error threshold;

training the additional type of machine learning algorithm using the training corpus to generate training results;

augmenting the machine learning model with the training results from the additional type of machine learning algorithm; and

determining that generation of the trained machine learning model is complete when additional training error measurement of the machine learning model following the augmenting is at or below the training error threshold.

17. The computer-implemented method of claim 16, further comprising refining the machine learning model based on user feedback regarding an effective of the solution for the root cause, the refining including retraining the machine learning model based on at least one of the training corpus that is modified based on the feedback or one or more modified algorithm selection rules.

18. The computer-implemented method of claim 16, wherein the performing the feature engineering includes performing at least one of outlier detection analysis to exclude outlier features in the training corpus from being use in generating the trained machine learning model, feature composition analysis to transform multiple features in the training corpus into a single feature, or feature relevance analysis to eliminate one or more redundant features from being used in generating the trained machine learning model.

19. The computer-implemented method of claim 16, wherein the initial type of machine learning algorithm or the additional type of machine learning algorithm is a Bayesian algorithm, a decision tree algorithm, a Support Vector Machine algorithm, an ensembles of trees algorithm, or an isotonic regression algorithm.

20. A system, comprising:

one or more processors; and

memory including a plurality of computer-executable components that are executable by the one or more processors to perform a plurality of actions, the plurality of actions comprising;

performing feature engineering on a training corpus for generating a machine learning model that is used to determine solutions for issues with a wireless carrier network;

training an initial type of machine learning algorithm using the training corpus to generate the machine learning model;

determining a training error measurement of the machine learning model is above a training error threshold;

selecting an additional type of machine learning algorithm based on a magnitude of the training error measurement according to one or more algorithm selection rules in response to the training error measurement being above the training error threshold;

training the additional type of machine learning algorithm using the training corpus to generate training results;

augmenting the machine learning model with the training results from the additional type of machine learning algorithm;

determining that generation of a trained machine learning model is complete when additional training error measurement of the machine learning model following the augmenting is at or below the training error threshold;

receiving performance data regarding user device and network components of the wireless carrier network from multiple data sources;

receiving an indication of an issue affecting one or more user devices that are using the wireless carrier network, the issue impacting a quality of service experienced by at least one user of the one or more user devices as the one or more user devices are used to make voice calls, make multimedia calls, upload data, or download data using the wireless carrier network;

analyzing the performance data using the trained machine learning model to determine a root cause for the issue affecting the one or more user devices, the trained machine learning model employing multiple types of machine learning algorithms to analyze the performance data; and

providing at least one of the root cause and a solution that resolves the root cause for presentation.