WSEAS Transactions on Computer Research
Print ISSN: 1991-8755, E-ISSN: 2415-1521
Volume 13, 2025
Handling Missing Data Techniques:
A Meta-Analysis
Author:
Abstract: The predictive performance of any classification or regression model highly depends on the quality of the collected data. Most of time datasets suffer from the problem of missing values, and hence, several techniques have been proposed to handle the problem of missing values. Consequently, this paper aims to quickly survey the most well-known techniques that handle missing data, and identify the best one to use concerning several issues such as the ratio of missing values, type of attributes in the dataset, number of instances, and number of class labels. Hence, seven different and well-known missing values handling techniques have been evaluated and compared using five datasets with different characteristics concerning the Accuracy metric. The results revealed that the K- Means technique is the most appropriate technique to handle the problem of missing data and the SMO classifier is the best choice to use as a classification model in case of missing data.
Search Articles
Pages: 178-186
DOI: 10.37394/232018.2025.13.18