Imputation of Missing Data with Class Imbalance using Conditional Generative Adversarial Networks

Missing data is a common problem faced with real-world datasets. Imputation\nis a widely used technique to estimate the missing data. State-of-the-art\nimputation approaches, such as Generative Adversarial Imputation Nets (GAIN),\nmodel the distribution of observed data to approximate the missing values. Such\nan approach usually models a single distribution for the entire dataset, which\noverlooks the class-specific characteristics of the data. Class-specific\ncharacteristics are especially useful when there is a class imbalance. We\npropose a new method for imputing missing data based on its class-specific\ncharacteristics by adapting the popular Conditional Generative Adversarial\nNetworks (CGAN). Our Conditional Generative Adversarial Imputation Network\n(CGAIN) imputes the missing data using class-specific distributions, which can\nproduce the best estimates for the missing values. We tested our approach on\nbenchmark datasets and achieved superior performance compared with the\nstate-of-the-art and popular imputation approaches.\n