Enhancing Machine Learning Robustness in Noisy and Incomplete Datasets
DOI:
https://doi.org/10.65150/EP-gjetr/V2E4/2026-03Keywords:
Datasets, Machine Learning, Noisy, Robustness, Artificial Intelligence and IncompleteAbstract
Machine learning (ML) models have achieved impressive performance across diverse applications. However, their accuracy and generalizability are often compromised by real-world data irregularities such as noise, outliers, and missing values. These imperfections are common in domains like healthcare, finance and remote sensing, where acquiring clean, complete datasets is challenging. This paper adopted a hybrid methodological approach comprising both experimental and analytical frameworks and to investigate techniques that improve the robustness of ML models when faced with such noisy or incomplete datasets. The results of the UCI Adult dataset show how sensitive machine learning models are to missing data under the Missing At Random (MAR) mechanism at a corruption threshold of thirty percent. When mean imputation and XGBoost were used together, the accuracy significantly decreased from 85% to 70%, or 15 percentage points (pp). This significant deterioration shows that basic imputation methods, such as mean replacement, are unable to maintain the underlying data distribution and variable linkages. The predictive power of models is ultimately weakened by mean imputation, which tends to increase bias and minimize variance. Specifically, it will explore methods like noise-tolerant training algorithms, imputation strategies, ensemble learning, and adversarial robustness. By doing so, the study contributes to the growing field of reliable and explainable artificial intelligence, ensuring ML systems remain effective in uncertain environments. In conclusion, this study underscores the need for developing machine learning models that remain dependable in the face of noisy and incomplete datasets and recommended amongst others that future ML development workflows integrate noise-handling techniques from the early stages of data preprocessing to model deployment.
References
1) Azur, M. J., Stuart, E. A., Frangakis, C., and Leaf, P. J. (2011). Multiple imputation by chained equations: what is it and how does it work?. International journal of methods in psychiatric research, 20(1), 40–49.
2) Batista, G. E. A. P. A., & Monard, M. C. (2003). An analysis of four missing data treatment methods for supervised learning. Applied Artificial Intelligence, 17(5–6), 519–533.
3) Dietterich, T. G. (2000). Ensemble methods in machine learning. In International workshop on multiple classifier systems (pp. 1–15). Springer.
4) Durbin, J., & Koopman, S. J. (2012). Time series analysis by state space methods (2nd ed.). Oxford University Press.
5) Gal, Y., and Ghahramani, Z. (2016). Dropout as a Bayesian approximation: Representing model uncertainty in deep learning. In International conference on machine learning (pp. 1050–1059).
6) García-Laencina, P. J., Sancho-Gómez, J. L., & Figueiras-Vidal, A. R. (2010). Pattern classification with missing data: A review. Neural Computing and Applications, 19(2), 263–282.
7) Goodfellow, I. J., Shlens, J., and Szegedy, C. (2015). Explaining and harnessing adversarial examples. In International Conference on Learning Representations (ICLR).
8) Han, J., Kamber, M., & Pei, J. (2012). Data mining: Concepts and techniques (3rd ed.). Morgan Kaufmann.
9) Kingma, D. P., & Welling, M. (2014). Auto-encoding variational Bayes. International Conference on Learning Representations (ICLR).
10) Kotsiantis, S. B., Zaharakis, I., & Pintelas, P. (2006). Machine learning: A review of classification and combining techniques. Artificial Intelligence Review, 26(3), 159–190.
11) Kotu, V., & Deshpande, B. (2019). Data science: Concepts and practice (2nd ed.). Morgan Kaufmann.
12) Little, R. J. A., & Rubin, D. B. (2002). Statistical analysis with missing data (2nd ed.). Wiley.
13) Nguyen, H., Tran, M. T., and Nguyen, M. T. (2020). A survey on outlier detection methods in streaming data. ACM Computing Surveys (CSUR), 53(3), 1–37.
14) Rubin, D. B. (1976). Inference and missing data. Biometrika, 63(3), 581–592.
15) Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state of the art. Psychological Methods, 7(2), 147–177.
16) Shorten, C., and Khoshgoftaar, T. M. (2019). A survey on image data augmentation for deep learning. Journal of Big Data, 6(1), 1–48.
17) Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., and Salakhutdinov, R. (2014). Dropout: a simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 15(1), 1929–1958.
18) Troyanskaya, O., Cantor, M., Sherlock, G., Brown, P., Hastie, T., Tibshirani, R., Botstein, D., & Altman, R. B. (2001). Missing value estimation methods for DNA microarrays. Bioinformatics, 17(6), 520–525.
19) Vincent, P., Larochelle, H., Bengio, Y., & Manzagol, P. A. (2008). Extracting and composing robust features with denoising autoencoders. Proceedings of ICML, 1096–1103.
20) Zhou, Z. H. (2012). Ensemble methods: Foundations and algorithms. CRC Press.
21) Zhu, X., and Wu, X. (2004). Class noise vs. attribute noise: A quantitative study. Artificial Intelligence Review, 22(3), 177–210.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Maduabuchukwu Christopher, Ekuerhare Okeraghogho (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.
