Missing Features Reconstruction Using a Wasserstein Generative Adversarial Imputation Network

TL;DR

This paper compares generative and non-generative models for missing feature reconstruction, introducing WGAIN, a Wasserstein-based GAIN variant, which outperforms other methods when missing data is up to 30%.

Contribution

The paper introduces WGAIN, a Wasserstein modification of GAIN, demonstrating its superior performance in missing data imputation for continuous features.

Findings

WGAIN outperforms other models when missingness ≤ 30%.

GAIN and WGAIN are the best imputation methods overall.

WGAIN consistently outperforms or matches traditional methods.

Abstract

Missing data is one of the most common preprocessing problems. In this paper, we experimentally research the use of generative and non-generative models for feature reconstruction. Variational Autoencoder with Arbitrary Conditioning (VAEAC) and Generative Adversarial Imputation Network (GAIN) were researched as representatives of generative models, while the denoising autoencoder (DAE) represented non-generative models. Performance of the models is compared to traditional methods k-nearest neighbors (k-NN) and Multiple Imputation by Chained Equations (MICE). Moreover, we introduce WGAIN as the Wasserstein modification of GAIN, which turns out to be the best imputation model when the degree of missingness is less than or equal to 30%. Experiments were performed on real-world and artificial datasets with continuous features where different percentages of features, varying from 10% to 50%, were missing. Evaluation of algorithms was done by measuring the accuracy of the classification model previously trained on the uncorrupted dataset. The results show that GAIN and especially WGAIN are the best imputers regardless of the conditions. In general, they outperform or are comparative to MICE, k-NN, DAE, and VAEAC.