Effect of Missing Data Types and Imputation Methods on Supervised Classifiers: An Evaluation Study
Data completeness is one of the most common challenges that hinder the performance of data analytics platforms. Different studies have assessed the effect of missing values on different classification models based on a single evaluation metric, namely, accuracy. However, accuracy on its own is a mis...
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2023-03-01
|
Series: | Big Data and Cognitive Computing |
Subjects: | |
Online Access: | https://www.mdpi.com/2504-2289/7/1/55 |
_version_ | 1797613528299864064 |
---|---|
author | Menna Ibrahim Gabr Yehia Mostafa Helmy Doaa Saad Elzanfaly |
author_facet | Menna Ibrahim Gabr Yehia Mostafa Helmy Doaa Saad Elzanfaly |
author_sort | Menna Ibrahim Gabr |
collection | DOAJ |
description | Data completeness is one of the most common challenges that hinder the performance of data analytics platforms. Different studies have assessed the effect of missing values on different classification models based on a single evaluation metric, namely, accuracy. However, accuracy on its own is a misleading measure of classifier performance because it does not consider unbalanced datasets. This paper presents an experimental study that assesses the effect of incomplete datasets on the performance of five classification models. The analysis was conducted with different ratios of missing values in six datasets that vary in size, type, and balance. Moreover, for unbiased analysis, the performance of the classifiers was measured using three different metrics, namely, the Matthews correlation coefficient (MCC), the F1-score, and accuracy. The results show that the sensitivity of the supervised classifiers to missing data differs according to a set of factors. The most significant factor is the missing data pattern and ratio, followed by the imputation method, and then the type, size, and balance of the dataset. The sensitivity of the classifiers when data are missing due to the Missing Completely At Random (MCAR) pattern is less than their sensitivity when data are missing due to the Missing Not At Random (MNAR) pattern. Furthermore, using the MCC as an evaluation measure better reflects the variation in the sensitivity of the classifiers to the missing data. |
first_indexed | 2024-03-11T06:56:02Z |
format | Article |
id | doaj.art-fa8a3367893f41e0824de55856259cd6 |
institution | Directory Open Access Journal |
issn | 2504-2289 |
language | English |
last_indexed | 2024-03-11T06:56:02Z |
publishDate | 2023-03-01 |
publisher | MDPI AG |
record_format | Article |
series | Big Data and Cognitive Computing |
spelling | doaj.art-fa8a3367893f41e0824de55856259cd62023-11-17T09:37:19ZengMDPI AGBig Data and Cognitive Computing2504-22892023-03-01715510.3390/bdcc7010055Effect of Missing Data Types and Imputation Methods on Supervised Classifiers: An Evaluation StudyMenna Ibrahim Gabr0Yehia Mostafa Helmy1Doaa Saad Elzanfaly2Department of Business Information Systems (BIS), Faculty of Commerce and Business Administration, Helwan University, Cairo 11795, EgyptDepartment of Business Information Systems (BIS), Faculty of Commerce and Business Administration, Helwan University, Cairo 11795, EgyptDepartment of Information Systems, Faculty of Computer and Artificial Intelligence, Helwan University, Cairo 11795, EgyptData completeness is one of the most common challenges that hinder the performance of data analytics platforms. Different studies have assessed the effect of missing values on different classification models based on a single evaluation metric, namely, accuracy. However, accuracy on its own is a misleading measure of classifier performance because it does not consider unbalanced datasets. This paper presents an experimental study that assesses the effect of incomplete datasets on the performance of five classification models. The analysis was conducted with different ratios of missing values in six datasets that vary in size, type, and balance. Moreover, for unbiased analysis, the performance of the classifiers was measured using three different metrics, namely, the Matthews correlation coefficient (MCC), the F1-score, and accuracy. The results show that the sensitivity of the supervised classifiers to missing data differs according to a set of factors. The most significant factor is the missing data pattern and ratio, followed by the imputation method, and then the type, size, and balance of the dataset. The sensitivity of the classifiers when data are missing due to the Missing Completely At Random (MCAR) pattern is less than their sensitivity when data are missing due to the Missing Not At Random (MNAR) pattern. Furthermore, using the MCC as an evaluation measure better reflects the variation in the sensitivity of the classifiers to the missing data.https://www.mdpi.com/2504-2289/7/1/55data qualitydata completenessmissing patternsimputation techniquessupervisedclassifiers |
spellingShingle | Menna Ibrahim Gabr Yehia Mostafa Helmy Doaa Saad Elzanfaly Effect of Missing Data Types and Imputation Methods on Supervised Classifiers: An Evaluation Study Big Data and Cognitive Computing data quality data completeness missing patterns imputation techniques supervised classifiers |
title | Effect of Missing Data Types and Imputation Methods on Supervised Classifiers: An Evaluation Study |
title_full | Effect of Missing Data Types and Imputation Methods on Supervised Classifiers: An Evaluation Study |
title_fullStr | Effect of Missing Data Types and Imputation Methods on Supervised Classifiers: An Evaluation Study |
title_full_unstemmed | Effect of Missing Data Types and Imputation Methods on Supervised Classifiers: An Evaluation Study |
title_short | Effect of Missing Data Types and Imputation Methods on Supervised Classifiers: An Evaluation Study |
title_sort | effect of missing data types and imputation methods on supervised classifiers an evaluation study |
topic | data quality data completeness missing patterns imputation techniques supervised classifiers |
url | https://www.mdpi.com/2504-2289/7/1/55 |
work_keys_str_mv | AT mennaibrahimgabr effectofmissingdatatypesandimputationmethodsonsupervisedclassifiersanevaluationstudy AT yehiamostafahelmy effectofmissingdatatypesandimputationmethodsonsupervisedclassifiersanevaluationstudy AT doaasaadelzanfaly effectofmissingdatatypesandimputationmethodsonsupervisedclassifiersanevaluationstudy |