GroupSegment-SHAP: Shapley Value Explanations with Group-Segment Players for Multivariate Time Series

TL;DR

This paper tackles the challenge of explaining multivariate time-series predictions by addressing the fragmentation of signals when separators for features and time are treated separately. It introduces GS-SHAP, which constructs group-segment players by (i) clustering variables with HSIC-based nonlinear dependence into groups and (ii) segmenting time via MMD-driven change-point detection, producing joint spatiotemporal explanatory units. Shapley values are then computed over these units with permutation-based masking, yielding faithful attributions that better recover the model's reasoning, as evidenced by an average gain of about $1.7\times$ in deletion-based $\Delta$AUC and a $40\%$ reduction in runtime relative to baselines under matched budgets. Across HAR, ETTm1, PTB-XL, and S&P500, GS-SHAP demonstrates stronger faithfulness and robustness, and a finance case study demonstrates regime-dependent multivariate temporal interactions, underscoring the method’s practical impact for interpretable decision-making in healthcare, energy, and finance. Future work aims to broaden applicability to more architectures and enhance Shapley estimation efficiency for longer time horizons and higher dimensional data.

Abstract

Multivariate time-series models achieve strong predictive performance in healthcare, industry, energy, and finance, but how they combine cross-variable interactions with temporal dynamics remains unclear. SHapley Additive exPlanations (SHAP) are widely used for interpretation. However, existing time-series variants typically treat the feature and time axes independently, fragmenting structural signals formed jointly by multiple variables over specific intervals. We propose GroupSegment SHAP (GS-SHAP), which constructs explanatory units as group-segment players based on cross-variable dependence and distribution shifts over time, and then quantifies each unit's contribution via Shapley attribution. We evaluate GS-SHAP across four real-world domains: human activity recognition, power-system forecasting, medical signal analysis, and financial time series, and compare it with KernelSHAP, TimeSHAP, SequenceSHAP, WindowSHAP, and TSHAP. GS-SHAP improves deletion-based faithfulness (DeltaAUC) by about 1.7x on average over time-series SHAP baselines, while reducing wall-clock runtime by about 40 percent on average under matched perturbation budgets. A financial case study shows that GS-SHAP identifies interpretable multivariate-temporal interactions among key market variables during high-volatility regimes.

Figures (8)

• Figure 1: Comparison of player designs in existing sequential SHAP variants and GS-SHAP.
• Figure 2: Overview of GS-SHAP Framework.
• Figure 3: Deletion curves across the four datasets.
• Figure 4: Comparison of feature grouping strategies.
• Figure 5: Cosine similarity under background changes.