Home // BIOTECHNO 2014, The Sixth International Conference on Bioinformatics, Biocomputational Systems and Biotechnologies // View article

Evaluation of Imputation Methods for Missing Data and Their Effect on the Reliability of Predictive Models

Authors:
Xiao-Ou Ping
Feipei Lai
Yi-Ju Tseng
Ja-Der Liang
Guan-Tarn Huang
Pei-Ming Yang

Keywords: incomplete data; missing value; predictive model; liver cancer

Abstract:
In medical research, the problem of missing data occurs frequently. In this paper, eight imputation methods are evaluated based on accuracy and stability through a simulation experiment. The objective of this paper is to find appropriate methods for handling incomplete data sets during the development of predictive models which predict the recurrence status of liver cancer patients. Support vector machine (SVM) is employed for building predictive models. The data sources produced by different missing data handling methods (complete variable analysis and imputation method) are used for evaluating the impact on the development of the recurrence predictive model. Imputation methods show the potential benefit of features with missing values during the development of the recurrence predictive model.

Pages: 8 to 14

Copyright: Copyright (c) IARIA, 2014

Publication date: April 20, 2014

Published in: conference

ISSN: 2308-4383

ISBN: 978-1-61208-335-3

Location: Chamonix, France

Dates: from April 20, 2014 to April 24, 2014