Histogram approaches for imbalanced data streams regression

2025-01-29

AI Summary

In data stream regression, rare samples are unevenly distributed across the entire target range—not merely at extremes—causing severe distributional imbalance that degrades model performance. Method: This paper proposes a histogram-driven online resampling framework, introducing HistUS (Histogram-based Online Undersampling) and HistOS (Histogram-based Online Oversampling). Unlike Chebyshev inequality–based approaches, which poorly localize rare instances, our methods employ dynamic binning to achieve adaptive rebalancing across the full target distribution. Integrated into a streaming architecture, the framework enables real-time adaptation to concept drift and sudden imbalance shifts. Results: Evaluated on multiple synthetic and real-world data streams, the approach reduces average MAE by 18.7% over state-of-the-art baselines, demonstrating significant improvements in both predictive accuracy and robustness to distributional changes.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Data Mining & Knowledge Management: Data Stream Mining
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications

Abstract

Handling imbalanced data streams in regression tasks presents a significant challenge, as rare instances can appear anywhere in the target distribution rather than being confined to its extreme values. In this paper, we introduce novel data-level sampling strategies, exttt{HistUS} and exttt{HistOS}, that utilize histogram-based approaches to dynamically balance data streams. Unlike previous methods based on Chebyshev extquotesingle s inequality, our proposed techniques identify and handle rare cases across the entire distribution effectively. We demonstrate that exttt{HistUS} and exttt{HistOS} outperform traditional methods through extensive experiments on synthetic and real-world datasets, leading to more accurate and robust regression models in streaming environments.

Problem

• Imbalanced Data
• Predictive Accuracy
• Flow Information Forecasting

Innovation

• Histogram Balancing
• Data Imbalance Solution
• Enhanced Prediction Accuracy