Fraud detection using emotion-based deep learning model

1. A computing system comprising:

a memory; and

one or more processors in communication with the memory and configured to:

receive a set of emotion factor values for communication data of a current communication associated with a customer, wherein each emotion factor value of the set of emotion factor values indicates a measure of a different emotion in the current communication;

classify, using an emotion variance model running on the one or more processors, the current communication into an emotional fraud category based on the set of emotion factor values for the current communication associated with the customer, wherein the emotion variance model comprises a machine learning model;

determine a risk score for the current communication indicative of a probability that the current communication is fraudulent based on at least the emotional fraud category for the current communication;

determine an accuracy of the emotion variance model based on example sets of emotion factor values corresponding to known emotional fraud categories; and

based on the accuracy not exceeding a threshold accuracy value, re-train the emotion variance model based on an updated set of training data, wherein the updated set of training data includes an updated plurality of customer communications, wherein each customer communication in the updated plurality of customer communications comprises a set of emotion factor values and a label indicating an emotional fraud category for the communication, and wherein the updated plurality of customer communications includes the current communication.

2. The computing system of claim 1 , wherein the one or more processors are further configured to:

receive the communication data of the current communication;

apply the communication data to an emotion-based indexer as input, wherein the emotion-based indexer includes a set of machine learning models, each machine learning model trained to determine the measure of the particular emotion factor in the current communication;

generate, as output from the emotion-based indexer, the set of emotion factor values for the current communication; and

store the set of emotion factor values for the current communication in a database.

3. The computing system of claim 1 , wherein the one or more processors are configured to classify the current communication into the emotional fraud category based on the set of emotion factor values for the current communication and one or more historic sets of emotion factor values stored in a database, wherein the one or more historic sets of emotion factor values correspond to communication data of one or more historic communications associated with the customer over time, the historic communications occurring prior to the current communication.

4. The computing system of claim 3 , wherein to classify the current communication, the one or more processors are configured to:

compute an average set of emotion factor values from the historic sets of emotion factor values for the historic communications associated with the customer over time; and

classify the current communication into the emotional fraud category based on the average set of emotion factor values for the customer over time.

5. The computing system of claim 4 , wherein the one or more processors are further configured to:

determine a set of delta values for the set of emotion factor values by calculating a difference between each emotion factor value of the set of emotion factor values for the current communication and each corresponding emotion factor value of the average set of emotion factor values for the historic communications; and

classify the current communication into the emotional fraud category in response to one or more delta values of the set of delta values exceeding a threshold.

6. The computing system of claim 4 , wherein the one or more processors are further configured to:

determine at least one delta value for at least one emotion factor value of the set of emotion factor values by calculating a difference between a particular emotion factor value for the current communication and a corresponding particular emotion factor value of the average set of emotion factor values for the historic communications; and

classify the current communication into the emotional fraud category based on the delta value for the particular emotion factor value exceeding a threshold.

7. The computing system of claim 5 , wherein to classify the communication, the one or more processors are configured to:

apply the set of delta values for the set of emotion factor values for the current communication to the emotion variance model as input; and

indicate, as output from the emotion variance model, the emotional fraud category for the current communication.

8. The computing system of claim 7 , wherein the one or more processors are configured to:

create a first set of training data that includes a first plurality of customer communications, wherein each customer communication in the first plurality of customer communications comprises a set of delta values and a label indicating an emotional fraud category for the communication; and

train the machine learning model based on the first set of training data.

9. The computing system of claim 1 , wherein to classify the communication, the one or more processors are configured to:

apply the set of emotion factor values for the current communication to the emotion variance model as input; and

indicate, as output from the emotion variance model, the emotional fraud category for the customer.

10. The computing system of claim 1 , wherein the one or more processors are configured to:

create a first set of training data that includes a first plurality of customer communications, wherein each customer communication in the first plurality of customer communications comprises a set of emotion factor values and a label indicating an emotional fraud category for the communication; and

train the machine learning model based on the first set of training data.

11. The computing system of claim 1 , wherein to classify the current communication, the one or more processors are configured to:

determine a pattern set of emotion factor values that corresponds to fraudulent communications;

determine a similarity between the pattern set of emotion factor values and the set of emotion factor values for the current communication; and

classify the current communication into the emotional fraud category based on the similarity.

12. The computing system of claim 1 , wherein to determine the risk score for the current communication, the one or more processors are further configured to:

receive a risk assessment of the current communication from a preliminary fraud detection system; and

determine the risk score for the current communication based on at least the risk assessment and the emotional fraud category for the current communication.

13. The computing system of claim 1 , wherein the one or more processors are further configured to transmit the risk score to one or more agent devices for use in determining how to handle the current communication.

14. The computing system of claim 1 , wherein the updated set of training data comprises a first updated set of training data, wherein the updated plurality of customer communications comprises a first plurality of customer communications, and wherein the one or more processors are configured to periodically re-train the emotion variance model based on a second updated set of training data, wherein the second updated set of training data includes a second updated plurality of customer communications including customer communications received since the emotion variance model was last trained, wherein each customer communication in the second updated plurality of customer communications comprises a set of emotion factor values and a label indicating an emotional fraud category for the communication.

15. A method comprising:

receiving, by one or more processors, a set of emotion factor values for communication data of a current communication associated with a customer, wherein each emotion factor value of the set of emotion factor values indicates a measure of a different emotion in the current communication;

classifying, using an emotion variance model running on the one or more processors, the current communication into an emotional fraud category based on the set of emotion factor values for the current communication associated with the customer, wherein the emotion variance model comprises a machine learning model;

determining a risk score for the current communication indicative of a probability that the current communication is fraudulent based on at least the emotional fraud category for the current communication;

determining an accuracy of the emotion variance model based on example sets of emotion factor values corresponding to known emotional fraud categories; and

based on the accuracy not exceeding a threshold accuracy value, re-training the emotion variance model based on an updated set of training data, wherein the updated set of training data includes an updated plurality of customer communications, wherein each customer communication in the updated plurality of customer communications comprises a set of emotion factor values and a label indicating an emotional fraud category for the communication, and wherein the updated plurality of customer communications includes the current communication.

16. The method of claim 15 , wherein classifying the current communication into the emotional fraud category is further based on the set of emotion factor values for the current communication and one or more historic sets of emotion factor values stored in a database, wherein the one or more historic sets of emotion factor values correspond to communication data of one or more historic communications associated with the customer over time, the historic communications occurring prior to the current communication.

17. The method of claim 16 , wherein classifying the current communication further comprises:

computing an average set of emotion factor values from the historic sets of emotion factor values for the historic communications associated with the customer over time;

determining a set of delta values for the set of emotion factor values by calculating a difference between each emotion factor value of the set of emotion factor values for the current communication and each corresponding emotion factor value of the average set of emotion factor values for the historic communications; and

classifying the current communication into the emotional fraud category in response to one or more delta values of the set of delta values exceeding a threshold.

18. The method of claim 17 , wherein classifying the current communication further comprises:

applying the set of delta values for the set of emotion factor values for the current communication to the emotion variance model as input; and

indicating, as output from the emotion variance model, the emotional fraud category for the customer.

19. The method of claim 18 further comprising:

creating a first set of training data that includes a first plurality of customer communications, wherein each customer communication in the first plurality of customer communications comprises a set of delta values and a label indicating an emotional fraud category for the customer communication; and

training the machine learning model based on the set of training data.

20. The method of claim 15 wherein classifying the communication further comprises:

applying the set of emotion factor values for the current communication to the emotion variance model as input; and

indicating, as output from the emotion variance model, the emotional fraud category for the customer.

21. The method of claim 15 , further comprising:

creating a first set of training data that includes a first plurality of customer communications, wherein each customer communication in the first plurality of customer communications comprises a set of emotion factor values and a label indicating an emotional fraud category for the communication; and

training the machine learning model based on the first set of training data.

22. A non-transitory computer-readable medium comprising instructions that, when executed, cause one or more processors to:

receive a set of emotion factor values for communication data of a current communication associated with a customer, wherein each emotion factor value of the set of emotion factor values indicates a measure of a different emotion in the current communication;

classify, using an emotion variance model running on the one or more processors, the current communication into an emotional fraud category based on the set of emotion factor values for the current communication associated with the customer, wherein the emotion variance model comprises a machine learning model;

determine a risk score for the current communication indicative of a probability that the current communication is fraudulent based on at least the emotional fraud category for the current communication;

determine an accuracy of the emotion variance model based on example sets of emotion factor values corresponding to known emotional fraud categories; and

based on the accuracy not exceeding a threshold accuracy value, re-train the emotion variance model based on an updated set of training data, wherein the updated set of training data includes an updated plurality of customer communications, wherein each customer communication in the updated plurality of customer communications comprises a set of emotion factor values and a label indicating an emotional fraud category for the communication, and wherein the updated plurality of customer communications includes the current communication.