Comparison of the Most Influential Missing Data Imputation Algorithms for Healthcare
Authors:
- 1 November 2018
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 247-251
Abstract
In healthcare research, the reliability of input data is essential. However, missing data is a common incident in this field for various reasons. Current research mainly focuses on developing new data imputation methodologies, while there is a need for studying on a global evaluation of existing algorithms. In this research, we compared the performance of four influential missing data imputation algorithms, Regularized Expectation-Maximization (EM), Multiple Imputation (MI), kNN Imputation (kNNI) and Mean Imputation on two real health care datasets: (1) MHEALTH dataset and (2) the University of Queensland Vital Signs dataset. Under the Missing Completely At Random (MCAR) assumption, Root Mean Squared Error (RMSE) and execution time were used as best performing evaluation criteria. The experimental analysis suggests that EM is the imputation algorithm which is expected to be a good choice to deal with the problem of missing data in the healthcare area.Keywords
This publication has 6 references indexed in Scilit:
- Tutorial on practical tips of the most influential data preprocessing algorithms in data miningKnowledge-Based Systems, 2016
- A Review of Missing Data Handling Methods in Education ResearchReview of Educational Research, 2014
- University of Queensland Vital Signs DatasetAnesthesia & Analgesia, 2012
- Review: A gentle introduction to imputation of missing valuesJournal of Clinical Epidemiology, 2006
- Maximizing the Usefulness of Data Obtained with Planned Missing Value Patterns: An Application of Maximum Likelihood ProceduresMultivariate Behavioral Research, 1996
- Maximum Likelihood from Incomplete Data Via the EM AlgorithmJournal of the Royal Statistical Society Series B: Statistical Methodology, 1977