Method, medium, and system for generating synthetic data

A method, comprising:

• categorizing historical consumer data based on a set of characteristics comprising at least one of transactional volume, a click-through rate, or whether the historical consumer data falls below a sparsity limit;
• receiving a first request to generate a first synthetic dataset, the first request specifying a first requirement for at least one of the characteristics, the first requirement identifying at least one of a plurality of categories;
• receiving a first request to generate a first synthetic dataset as training data to a machine learning system, the first request specifying a first requirement for at least one of the characteristics, the first requirement identifying at least one of the plurality of categories;
• providing a first subset of the historical consumer data as input to a data model to generate the first synthetic dataset, the data model mapping from a random or pseudorandom vector to elements in a training data space, wherein generating the first synthetic dataset comprises determining a vector connecting a first representative point and a second representative point in code space;
• and
• providing the first synthetic dataset as training data to a machine learning system.

A non-transitory memory storing instructions that, when executed by at least one processor, cause a system to perform operations comprising:

• categorizing historical consumer data based on a set of characteristics comprising at least one of transactional volume, click-through rate, or whether the historical consumer data falls below a sparsity limit;
• receiving a first request to generate a first synthetic dataset, the first request specifying a first requirement for at least one of the characteristics, the first requirement identifying at least one of a plurality of categories;
• retrieving, from the historical consumer data, a first subset of the historical consumer data satisfying the first requirement;
• providing the first subset of the historical consumer data as input to a data model to generate the first synthetic dataset, the data model mapping from a random or pseudorandom vector to elements in a training data space, wherein generating the first synthetic dataset comprises determining a vector connecting a first representative point and a second representative point in code space;
• and
• providing the first synthetic dataset as training data to a machine learning system to determine at least one of a consumer behavior or a purchase recommendation.