Description: Synthetic data involves data generated artificially instead of from real operations. Synthetic data can bridge information silos by serving as a proxy for real data and not giving away sensitive information, for example, about people and intellectual property. Synthetic data generation capabilities can address many specific AI and analytics concerns (such as data scarcity, accessibility, bias, privacy and regulatory compliance). Data can be generated using different methods. These might include statistically rigorous sampling from real data, semantic approaches or generative adversarial networks, or by creating simulation scenarios where models and processes interact to create completely new datasets of events. Tabular synthetic data is in the form of data records, sometimes known as “structured” data, as opposed to image/video or language data. Tabular synthetic data includes longitudinal data and time-series data, plus it is the most relevant form of synthetic data for data and analytics. Uses for tabular synthetic data are to:
Train ML models
■
Test systems of many kinds
■
Build analytics sandboxes for exploration
■
Demonstrate data products with easy-to-create, yet realistic looking data (using a company’s real product list, for example), but simulating transactions ■
Sample Vendors: MOSTLY AI; Simudyne; Anonos (Statice); Tonic.ai
Range: Midrange (3 to 6 Years)
In insurance, a lack of joined up datasets across BUs and subsidiaries, as well as functional silos (claims, underwriting) is a real problem for insurers in training their models. Tabular synthetic data is still three to six years from early adoption. Many attempts at joining up data for AI and analytics projects will fail due to lack of progress with data warehousing and an inability to extract data from legacy storage. Statistical issues such as left truncation from uneven deductibles or loss runs at different layers create issues in using historical claims data for actuarial modeling, too.
Gartner, Inc. | G00786204
Page 26 of 48
This research note is restricted to the personal use of abhishek.sharma@fractal.ai.
Made with FlippingBook - PDF hosting