Predictive model for generating paired identifiers

1. A computer-implemented method comprising:

accessing, by one or more data processors, a training dataset comprising a plurality of data pairs, wherein each data pair of the plurality of data pairs includes first data from a first recorded observation for a first dependent user and second data from a first charge code for the first dependent user, the first charge code corresponds to the first recorded observation, the first charge code comprises structured data, and a format of the first charge code is changed to generate the second data;

training, by the one or more data processors, a predictive model by using the training dataset;

accessing, by the one or more data processors, an unmatched recorded observation for a second dependent user, wherein the unmatched recorded observation is missing a corresponding charge code that is related to the first charge code;

counting, by the one or more data processors, a number of completed fields within the unmatched recorded observation;

at least one of:

using, by the one or more data processors, the number of completed fields within the unmatched recorded observation to identify at least one of a plurality of paired charge codes; or

using, by the one or more data processors, the number of completed fields within the unmatched recorded observation to determine a number of the plurality of paired charge codes;

using, by the one or more data processors, the trained predictive model to generate at least one paired charge code for the unmatched recorded observation; and

outputting, by the one or more data processors, the at least one paired charge code,

wherein the at least one paired charge code comprises the plurality of paired charge codes that are ranked in order of likelihood of acceptance for the unmatched recorded observation.

2. The computer-implemented method of claim 1 , wherein the first recorded observation comprises unstructured data, and the method further comprises using natural language processing to generate the first data from the first recorded observation as structured data.

3. The computer-implemented method of claim 1 , wherein each data pair of the plurality of data pairs further includes a tag indicating whether the first charge code was accepted in association with the first recorded observation.

4. The computer-implemented method of claim 1 , wherein for each data pair of the plurality of data pairs, the training dataset further comprises information from at least one record of the first dependent user.

5. The computer-implemented method of claim 1 ,

wherein the training dataset further comprises information from at least one third party source that processes charge codes for dependent users.

6. A system comprising:

one or more data processors; and

a non-transitory computer readable storage medium containing instructions that, when executed on the one or more data processors, cause the one or more data processors to perform actions including:

accessing a training dataset comprising a plurality of data pairs, wherein each data pair of the plurality of data pairs includes first data from a first recorded observation for a first dependent user and second data from a first charge code for the first dependent user, the first charge code corresponds to the first recorded observation, the first charge code comprises structured data, and a format of the first charge code is changed to generate the second data;

training a predictive model by using the training dataset;

accessing an unmatched recorded observation for a second dependent user, wherein the unmatched recorded observation is missing a corresponding charge code that is related to the first charge code;

counting a number of completed fields within the unmatched recorded observation;

at least one of:

using the number of completed fields within the unmatched recorded observation to identify at least one of a plurality of paired charge codes; or

using the number of completed fields within the unmatched recorded observation to determine a number of the plurality of paired charge codes;

using the trained predictive model to generate at least one paired charge code for the unmatched recorded observation; and

outputting the at least one paired charge code,

wherein the at least one paired charge code comprises the plurality of paired charge codes that are ranked in order of likelihood of acceptance for the unmatched recorded observation.

7. The system of claim 6 , wherein the first recorded observation comprises unstructured data, and the actions further include using natural language processing to generate the first data from the first recorded observation as structured data.

8. The system of claim 6 , wherein each data pair of the plurality of data pairs further includes a tag indicating whether the first charge code was accepted in association with the first recorded observation.

9. The system of claim 6 , wherein for each data pair of the plurality of data pairs, the training dataset further comprises information from at least one record of the first dependent user.

10. The system of claim 6 , wherein the training dataset further comprises information from at least one third party source that processes charge codes for dependent users.

11. A computer-program product tangibly embodied in a non-transitory machine-readable storage medium, including instructions configured to cause one or more data processors to perform actions including:

accessing a training dataset comprising a plurality of data pairs, wherein each data pair of the plurality of data pairs includes first data from a first recorded observation for a first dependent user and second data from a first charge code for the first dependent user, and the first charge code corresponds to the first recorded observation, the first charge code comprises structured data, and a format of the first charge code is changed to generate the second data;

training a predictive model by using the training dataset;

accessing an unmatched recorded observation for a second dependent user, wherein the unmatched recorded observation is missing a corresponding charge code that is related to the first charge code;

counting a number of completed fields within the unmatched recorded observation;

at least one of:

using the number of completed fields within the unmatched recorded observation to identify at least one of a plurality of paired charge codes; or

using the number of completed fields within the unmatched recorded observation to determine a number of the plurality of paired charge codes;

using the trained predictive model to generate at least one paired charge code for the unmatched recorded observation; and

outputting the at least one paired charge code,

wherein the at least one paired charge code comprises the plurality of paired charge codes that are ranked in order of likelihood of acceptance for the unmatched recorded observation.

12. The computer-program product of claim 11 , wherein the first recorded observation comprises unstructured data, and the actions further include using natural language processing to generate the first data from the first recorded observation as structured data.

13. The computer-program product of claim 11 , wherein each data pair of the plurality of data pairs further includes a tag indicating whether the first charge code was accepted in association with the first recorded observation.

14. The computer-program product of claim 11 , wherein for each data pair of the plurality of data pairs, the training dataset further comprises information from at least one record of the first dependent user.