![[2510.26046] Bias-Corrected Data Synthesis for Imbalanced Learning](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2510.26046] Bias-Corrected Data Synthesis for Imbalanced Learning
Summary
This paper presents a method for bias-corrected data synthesis aimed at improving classification accuracy in imbalanced learning scenarios by addressing the limitations of synthetic data generation.
Why It Matters
Imbalanced datasets are common in machine learning, often leading to poor model performance. This research provides a novel approach to mitigate bias in synthetic data, enhancing the reliability of predictive models, which is crucial for applications across various domains, including healthcare and finance.
Key Takeaways
- Synthetic data can introduce bias that negatively impacts model accuracy.
- The proposed bias correction procedure improves prediction accuracy and reduces overfitting.
- The method is applicable to various imbalanced learning scenarios, including multi-task learning.
Statistics > Machine Learning arXiv:2510.26046 (stat) [Submitted on 30 Oct 2025 (v1), last revised 15 Feb 2026 (this version, v2)] Title:Bias-Corrected Data Synthesis for Imbalanced Learning Authors:Pengfei Lyu, Zhengchi Ma, Linjun Zhang, Anru R. Zhang View a PDF of the paper titled Bias-Corrected Data Synthesis for Imbalanced Learning, by Pengfei Lyu and 2 other authors View PDF HTML (experimental) Abstract:Imbalanced data, where the positive samples represent only a small proportion compared to the negative samples, makes it challenging for classification problems to balance the false positive and false negative rates. A common approach to addressing the challenge involves generating synthetic data for the minority group and then training classification models with both observed and synthetic data. However, since the synthetic data depends on the observed data and fails to replicate the original data distribution accurately, prediction accuracy is reduced when the synthetic data is naïvely treated as the true data. In this paper, we address the bias introduced by synthetic data and provide consistent estimators for this bias by borrowing information from the majority group. We propose a bias correction procedure to mitigate the adverse effects of synthetic data, enhancing prediction accuracy while avoiding overfitting. This procedure is extended to broader scenarios with imbalanced data, such as imbalanced multi-task learning and causal inference. Theoretical properties,...