Using neural networks for imputing missing values in insurance data

Real data often contains missing values which poses a major challenge in predictive modeling. Imputation techniques are proposed to impute missing values. In this paper, we explore the use of neural networks to impute missing values. In particular, we extend and explore an imputation method based on generative adversarial networks (GANs) to impute missing values in insurance data. The imputation method can handle missing values in datasets that contain both categorical and continuous variables. We conduct experiments on large insurance datasets from different areas such as life insurance and property and casualty insurance. Our numerical results show that using the GAN-based imputation method to impute missing values is efficient and helps to improve the prediction accuracy.