IMFACT: Counterfactual Explanations for Time Series via Intrinsic Mode Function Substitution

2026-08-05

AI Summary

This work addresses the challenge that counterfactual perturbations applied directly in the original feature space of oscillatory time series often disrupt their temporal structure, yielding physically implausible explanations. To overcome this, the authors propose a model-agnostic counterfactual generation framework that first decomposes the input signal into intrinsic mode functions (IMFs) via empirical mode decomposition (EMD), then selects critical IMFs based on variance. A three-nearest unlike neighbor (NUN) iterative replacement strategy is employed to modify these key IMFs, guiding the classifier toward the target class while preserving the signal’s physical plausibility. This approach represents the first method to construct counterfactual explanations in the IMF space. Experiments on the FaultDetectionA and FruitFlies datasets demonstrate consistent improvements over existing baselines across reliability, plausibility, and proximity metrics.

Abstract

Oscillatory signals, such as vibration, carry class-discriminative information in specific frequency bands; perturbing them in raw feature space for counterfactual analysis easily destroys their temporal structure and produces physically implausible results. In this work, we introduce IMFACT (IMF-based counterfACTuals), a model-agnostic framework for generating plausible counterfactual explanations for time series classifiers that operates in the decomposition space of Empirical Mode Decomposition. An input signal is split into Intrinsic Mode Functions (IMFs), and selected IMFs are progressively substituted with those of a Nearest Unlike Neighbour (NUN) until the classifier flips to the target class. We evaluate six IMF-selection strategies and a multi-NUN cycling extension on two UCR benchmarks (FaultDetectionA, FruitFlies). The variance-based strategy with three NUNs outperforms two prominent baseline techniques on reliability and plausibility metrics, while cycling across three NUNs yields the best proximity across both datasets.

Problem

• counterfactual explanations
• time series
• temporal structure
• physically implausible
• oscillatory signals

Innovation

• counterfactual explanations
• Intrinsic Mode Functions
• Empirical Mode Decomposition
• time series classification
• model-agnostic