Establishing strong imputation performance of a denoising autoencoder in a wide range of missing data problems

Research output: Contribution to journalArticle

Abstract

Dealing with missing data in data analysis is inevitable. Although powerful imputation methods that address this problem exist, there is still much room for improvement. In this study, we examined single imputation based on deep autoencoders, motivated by the apparent success of deep learning to efficiently extract useful dataset features. We have developed a consistent framework for both training and imputation. Moreover, we benchmarked the results against state-of-the-art imputation methods on different data sizes and characteristics. The work was not limited to the one-type variable dataset; we also imputed missing data with multi-type variables, e.g., a combination of binary, categorical, and continuous attributes. To evaluate the imputation methods, we randomly corrupted the complete data, with varying degrees of corruption, and then compared the imputed and original values. In all experiments, the developed autoencoder obtained the smallest error for all ranges of initial data corruption.

Details

Authors
Organisations
External organisations
  • Halmstad University
Research areas and keywords

Subject classification (UKÄ) – MANDATORY

  • Bioinformatics (Computational Biology)

Keywords

  • Autoencoder, Deep learning, Imputation, Missing data
Original languageEnglish
Pages (from-to)137-146
Number of pages10
JournalNeurocomputing
Volume365
Early online date2019 Jul 26
Publication statusPublished - 2019 Nov 6
Publication categoryResearch
Peer-reviewedYes