Missing Data Imputation Techniques: An Error Rate Evaluation Across Numerical Datasets

View at publisher Summarise with VIRAT
Authors: Manuka Pinidiyaarchchi, Devaan D Walakumburage, Tilani Gunawardena
Cited by: 0
References: 9

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Missing data is a critical challenge in numerical datasets, affecting the reliability of data analysis in domains such as healthcare, meteorology, and sales. This study evaluates seven imputation techniques: Mean, Median, Mode, K-Nearest Neighbors (KNN), Autoencoder, Multiple Imputation by Chained Equations (MICE), and Extreme Gradient Boosting (XGBoost) across three publicly available datasets with artificially introduced missing values ranging from 10% to 50%. The performance of each technique is assessed using Root Mean Squared Error, Mean Absolute Error, and R2. Results indicate that machine learning based methods, particularly XGBoost and MICE, consistently outperform traditional statistical approaches, maintaining higher accuracy across different domains and missingness levels. These findings provide practical guidance for selecting imputation techniques that ensure data quality and reliable analysis.

Fields of Study & Subjects

Statistical Methods and Bayesian Inference Machine Learning and Data Classification Data Mining and Machine Learning Applications

Publication details

Journal / source
IEEE International Conference on Industrial and Information Systems (ICIIS)
Pages
25–30
Publication date
16 January 2026
Language
EN