Efficient Utilization of Missing Data in Cost-Sensitive Learning
- 28 November 2019
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Knowledge and Data Engineering
- Vol. 33 (6) , 2425-2436
Abstract
Different from previous imputation methods which impute the missing values in the incomplete samples by using the information on the complete samples, this paper proposes a Date-drive Incremental imputation Model, DIM for short, to impute missing values economically, effectively, orderly, and iteratively, using all available information to conduct missing value imputation. To do this, we propose a scoring rule to rank the missing features by taking into account both the economical criterion and the effective imputation information. The economical criterion combines the imputation cost with the discriminative ability of the feature together, while the effective imputation information enables our method to use all the observed information or even the missing value imputed before to impute the left missing values. During the imputation process, our DIM first detects the neednot-impute samples for reducing the imputation cost as well as noise, and then selects the missing features with the top rank to be imputed. The imputation process orderly imputes missing values until all missing values are imputed or the imputation cost is exhausted. Experimental results on UCI data sets demonstrated the advantages of our proposed DIM, compared to the comparison methods, in terms of prediction accuracy and classification accuracy.Keywords
Funding Information
- China Key Research Program (2016YFB1000905)
- Natural Science Foundation of China (61836016, 61876046, 61573270, 61672177)
- Project of Guangxi Science and Technology (GuiKeAD17195062)
- Research Fund of Guangxi Key Lab of Multisource Information Mining and Security (18-A-01-01)
This publication has 49 references indexed in Scilit:
- A web-based approach to data imputationWorld Wide Web, 2013
- Missing data analyses: a hybrid multiple imputation algorithm using Gray System Theory and entropy based on clusteringApplied Intelligence, 2013
- Decision tree classifiers sensitive to heterogeneous costsJournal of Systems and Software, 2012
- Shell-neighbor method and its application in missing data imputationApplied Intelligence, 2010
- Incremental Tree-Based Missing Data Imputation with Lexicographic OrderingJournal of Classification, 2009
- Active Feature-Value AcquisitionManagement Science, 2009
- Comparison of Two Approaches for Handling Missing Covariates in Logistic RegressionEducational and Psychological Measurement, 2007
- Bayesian networks for imputation in classification problemsJournal of Intelligent Information Systems, 2007
- The problem of disguised missing dataACM SIGKDD Explorations Newsletter, 2006
- The use of background knowledge in decision tree inductionMachine Learning, 1991