Systems and methods for using synthetic data to ensure training data is properly representative
AssigneeWells Fargo Bank, N.A.
InventorsVinothkumar Venkataraman, Yang Angelina Yang, Ashutosh Verma, Naveen Gururaja Yeri
Life of this patent
Estimating the term…
Abstract
Systems, apparatuses, methods, and computer program products are disclosed for generating representative training data. An example method includes comparing, by data analysis circuitry, a labeled dataset to a target dataset. The example method also includes generating, by training data circuitry, a training dataset based on the comparison, wherein the training dataset comprises at least a portion of the labeled dataset supplemented by a synthetic dataset. The example method also includes training, by modeling circuitry, a model using the training dataset.
Claims20
A method for training a model using representative training data, the method comprising:
- generating, by training data circuitry, a training dataset that is representative of a target dataset and comprises at least a portion of an authentic dataset supplemented by synthetic data, wherein generating the training dataset comprises:
- populating, by the training data circuitry, the training dataset with a first authentic data point of the authentic dataset based on the first authentic data point satisfying a similarity threshold with a data point of the target dataset,
- determining, by the training data circuitry, a second data point of the target dataset that does not satisfy a similarity threshold with any authentic data points of the authentic dataset,
- generating, by synthetic data generation circuitry, a synthetic data point representative of the second data point, and
- populating, by the training data circuitry, the training dataset with the synthetic data point;
- and
- training, by modeling circuitry, the model using the training dataset.
An apparatus for training a model using representative training data, the apparatus comprising:
- training data circuitry configured to generate a training dataset that is representative of a target dataset and comprises at least a portion of an authentic dataset supplemented by synthetic data by:
- populating the training dataset with a first authentic data point of the authentic dataset based on the first authentic data point satisfying a similarity threshold with a data point of the target dataset,
- determining, by the training data circuitry, a second data point of the target dataset that does not satisfy a similarity threshold with any authentic data points of the authentic dataset, and
- populating, by the training data circuitry, the training dataset with a synthetic data point representative of the second data point;
- synthetic data generation circuitry configured to generate the synthetic data point representative of the second data point, and
- modeling circuitry configured to train the model using the training dataset.
A computer program product for training a model using representative training data, the computer program product comprising at least one non-transitory computer-readable storage medium storing software instructions that, when executed, cause an apparatus to:
- generating, by training data circuitry, a training dataset that is representative of a target dataset and comprises at least a portion of an authentic dataset supplemented by synthetic data, wherein generating the training dataset comprises:
- populating, by the training data circuitry, the training dataset with a first authentic data point of the authentic dataset based on the first authentic data point satisfying a similarity threshold with a data point of the target dataset,
- determining, by the training data circuitry, a second data point of the target dataset that does not satisfy a similarity threshold with any authentic data points of the authentic dataset,
- generating, by synthetic data generation circuitry, a synthetic data point representative of the second data point, and
- populating, by the training data circuitry, the training dataset with the synthetic data point;
- and
- training, by modeling circuitry, the model using the training dataset.
Dependencies are inferred from the claim text.