Pipelined machine learning frameworks
In general, embodiments of the present invention provide methods, apparatus, systems, computing devices, computing entities, and/or the like for pipelined machine learning. Certain embodiments of the present invention utilize systems, methods, and computer program products that perform pipelined machine learning using at least one of error correction machine learning models and output augmentation machine learning models, for example using a pipelined implementation of error correction machine learning models followed by output augmentation machine learning models.
1. A computer-implemented method for generating a predictive output based at least in part on an input data object, the computer-implemented method comprising:
generating, by one or more processors, one or more inference error correction engineered features based at least in part on the input data object, wherein the one or more inference error correction engineered features include an agent-based error likelihood value describing an estimated error likelihood of an input provider agent associated with the input data object;
processing, by the one or more processors, the one or more inference error correction engineered features, using a trained error correction machine learning model of a pipelined machine learning framework, to generate an error-corrected input data object in which one or more input data fields corresponding to the input data object are adjusted based at least in part on the one or more inference error correction engineered features; and
generating, by the one or more processors, the predictive output based at least in part on the error-corrected input data object.
2. The computer-implemented method of claim 1 , wherein the one or more inference error correction engineered features comprise a per-field-type error likelihood value for each selected input data field of one or more selected input fields of the input data object that describes an estimated error likelihood of an input data field type associated with the selected input data field.
3. The computer-implemented method of claim 1 , wherein generating the trained error correction machine learning model comprises:
identifying one or more training error correction data objects;
for each training error correction data object of the one or more training error correction data objects:
determining a set of training error correction engineered features, and
determining a finalized training error correction data object based at least in part on the set of training error correction engineered features; and
generating the trained error correction machine learning model based at least in part on each finalized training error correction data object for a training error correction data object of the one or more training error correction data objects.
4. The computer-implemented method of claim 3 , wherein each training error correction data object of the one or more training error correction data objects may describe a per-field error likelihood value for each selected input data field of one or more selected input fields of the input data object.
5. The computer-implemented method of claim 3 , wherein generating the trained error correction machine learning model based at least in part on each finalized training error correction data object for the training error correction data object comprises performing cross-validation with grid search using each finalized training error correction data object for the training error correction data object to generate the trained error correction machine learning model.
6. The computer-implemented method of claim 3 , wherein determining the finalized training error correction data object comprises performing mean target feature encoding with smoothing on the training error correction data object to determine the finalized training error correction data object.
7. The computer-implemented method of claim 1 , wherein the trained error correction machine learning model comprises a trained gradient boosting machine learning model.
8. A system for generating a predictive output based at least in part on an input data object, the system comprising one or more processors and memory including program code, the memory and the program code configured to, with the one or more processors, cause the system to at least:
generate one or more inference error correction engineered features based at least in part on the input data object, wherein the one or more inference error correction engineered features include an agent-based error likelihood value describing an estimated error likelihood of an input provider agent associated with the input data object;
process the one or more inference error correction engineered features using a trained error correction machine learning model of a pipelined machine learning framework to generate an error-corrected input data object in which one or more input data fields corresponding to the input data object are adjusted based at least in part on the one or more inference error correction engineered features; and
generate the predictive output based at least in part on the error-corrected input data object.
9. The system of claim 8 , wherein the one or more inference error correction engineered features comprise a per-field-type error likelihood value for each selected input data field of one or more selected input fields of the input data object that describes an estimated error likelihood of an input data field type associated with the selected input data field.
10. The system of claim 8 , wherein generating the trained error correction machine learning model comprises:
identifying one or more training error correction data objects;
for each training error correction data object of the one or more training error correction data objects:
determining a set of training error correction engineered features, and
determining a finalized training error correction data object based at least in part on the set of training error correction engineered features; and
generating the trained error correction machine learning model based at least in part on each finalized training error correction data object for a training error correction data object of the one or more training error correction data objects.
11. The system of claim 10 , wherein each training error correction data object of the one or more training error correction data objects may describe a per-field error likelihood value for each selected input data field of one or more selected input fields of the input data object.
12. The system of claim 10 , wherein generating the trained error correction machine learning model based at least in part on each finalized training error correction data object for the training error correction data object comprises performing cross-validation with grid search using each finalized training error correction data object for the training error correction data object to generate the trained error correction machine learning model.
13. The system of claim 10 , wherein determining the finalized training error correction data object comprises performing mean target feature encoding with smoothing on the training error correction data object to determine the finalized training error correction data object.
14. The system of claim 8 , wherein the trained error correction machine learning model comprises a trained gradient boosting machine learning model.
15. A computer program product for generating a predictive output based at least in part on an input data object, the computer program product comprising at least one non-transitory computer-readable storage medium having computer-readable program code portions stored therein, the computer-readable program code portions configured to:
generate one or more inference error correction engineered features based at least in part on the input data object, wherein the one or more inference error correction engineered features include an agent-based error likelihood value describing an estimated error likelihood of an input provider agent associated with the input data object;
process the one or more inference error correction engineered features using a trained error correction machine learning model of a pipelined machine learning framework to generate an error-corrected input data object in which one or more input data fields corresponding to the input data object are adjusted based at least in part on the one or more inference error correction engineered features; and
generate the predictive output based at least in part on the error-corrected input data object.
16. The computer program product of claim 15 , wherein the one or more inference error correction engineered features comprise a per-field-type error likelihood value for each selected input data field of one or more selected input fields of the input data object that describes an estimated error likelihood of an input data field type associated with the selected input data field.
17. The computer program product of claim 15 , wherein generating the trained error correction machine learning model comprises:
identifying one or more training error correction data objects;
for each training error correction data object of the one or more training error correction data objects:
determining a set of training error correction engineered features, and
determining a finalized training error correction data object based at least in part on the set of training error correction engineered features; and
generating the trained error correction machine learning model based at least in part on each finalized training error correction data object for a training error correction data object of the one or more training error correction data objects.
18. The computer program product of claim 17 , wherein each training error correction data object of the one or more training error correction data objects may describe a per-field error likelihood value for each selected input data field of one or more selected input fields of the input data object.
19. The computer program product of claim 17 , wherein generating the trained error correction machine learning model based at least in part on each finalized training error correction data object for the training error correction data object comprises performing cross-validation with grid search using each finalized training error correction data object for the training error correction data object to generate the trained error correction machine learning model.
20. The computer program product of claim 17 , wherein determining the finalized training error correction data object comprises performing mean target feature encoding with smoothing on the training error correction data object to determine the finalized training error correction data object.