Generative Synthesis of Insurance Datasets

Abstract

One of the impediments in advancing actuarial research and developing open source assets for insurance analytics is the lack of realistic publicly available datasets. In this work, we develop a workflow for synthesizing insurance datasets leveraging CTGAN, a recently proposed neural network architecture for generating tabular data. Applying the proposed workflow to publicly available data in the domains of general insurance pricing and life insurance shock lapse modeling, we evaluate the synthesized datasets from a few perspectives: machine learning efficacy, distributions of variables, and stability of model parameters. This workflow is implemented via an R interface to promote adoption by researchers and data owners.