ICM Ensemble with Novel Betting Functions for Concept Drift

TL;DR

This work advances concept drift detection by strengthening Inductive Conformal Martingales (ICM) with a CAUTIOUS betting function that leverages multiple density estimators. It introduces two estimators—Interpolated Histogram and k-Nearest Neighbors density—and combines them in ensemble ICMs, enabling rapid CD detection and robust retraining while preserving prediction availability. Empirical results on synthetic (STAGGER, SEA) and real-world (ELEC, AIRLINES) datasets show improved accuracy relative to prior CAUTIOUS approaches and competitive performance against state-of-the-art CD detectors. The findings highlight practical benefits for streaming systems requiring probabilistic guarantees, fast adaptation, and reliable predictions under various drift regimes.

Abstract

This study builds upon our previous work by introducing a refined Inductive Conformal Martingale (ICM) approach for addressing Concept Drift (CD). Specifically, we enhance our previously proposed CAUTIOUS betting function to incorporate multiple density estimators for improving detection ability. We also combine this betting function with two base estimators that have not been previously utilized within the ICM framework: the Interpolated Histogram and Nearest Neighbor Density Estimators. We assess these extensions using both a single ICM and an ensemble of ICMs. For the latter, we conduct a comprehensive experimental investigation into the influence of the ensemble size on prediction accuracy and the number of available predictions. Our experimental results on four benchmark datasets demonstrate that the proposed approach surpasses our previous methodology in terms of performance while matching or in many cases exceeding that of three contemporary state-of-the-art techniques.

Figures (8)

• Figure 1: STAGGER dataset accuracy with R=10 and 10% noise using the MIHNN betting function and all possible combinations of 10 classifiers.
• Figure 2: STAGGER dataset's number of non-available predictions with R=10 and 10% noise using the MIHNN betting function and all possible combinations of 10 classifiers.
• Figure 3: SEA dataset accuracy with R=10 and 10% noise using the MIHNN betting function and all possible combinations of 10 classifiers.
• Figure 4: SEA dataset's number of non-available predictions with R=10 and 10% noise using the MIHNN betting function and all possible combinations of 10 classifiers.
• Figure 5: ELEC dataset accuracy with R=10 using the MIHNN betting function and all possible combinations of 10 classifiers.

Theorems & Definitions (16)

• Proposition 1
• proof : Proof of Proposition \ref{['prob:1']}
• Theorem 1
• proof : Proof of Theorem \ref{['thm:1']}
• Theorem 2
• proof : Proof of Theorem \ref{['thm:2']}
• Theorem 3
• proof : Proof of Theorem \ref{['thm:3']}
• Theorem 4
• proof : Proof of Theorem \ref{['thm:4']}