Data-Aware and Scalable Sensitivity Analysis for Decision Tree Ensembles
Namrita Varshney, Ashutosh Gupta, Arhaan Ahmad, Tanay V. Tayal, S. Akshay
TL;DR
This work tackles robustness and fairness in tree-ensemble classifiers by formalizing feature sensitivity and introducing data-aware sensitivity verification. It combines MILP and SMT techniques to efficiently verify sensitivity, extends the approach to multiclass ensembles via a data-aware objective that favors counterexamples near the training distribution, and introduces the SViM tool with optimizations that scale to large ensembles (up to 800 trees of depth 8). The method yields higher-quality counterexamples and substantial speedups over prior work, enabling practical reliability guarantees in high-stakes domains. The framework lays groundwork for future efforts to harden models during training and to extend data-awareness to broader verification tasks across tree-based models.
Abstract
Decision tree ensembles are widely used in critical domains, making robustness and sensitivity analysis essential to their trustworthiness. We study the feature sensitivity problem, which asks whether an ensemble is sensitive to a specified subset of features -- such as protected attributes -- whose manipulation can alter model predictions. Existing approaches often yield examples of sensitivity that lie far from the training distribution, limiting their interpretability and practical value. We propose a data-aware sensitivity framework that constrains the sensitive examples to remain close to the dataset, thereby producing realistic and interpretable evidence of model weaknesses. To this end, we develop novel techniques for data-aware search using a combination of mixed-integer linear programming (MILP) and satisfiability modulo theories (SMT) encodings. Our contributions are fourfold. First, we strengthen the NP-hardness result for sensitivity verification, showing it holds even for trees of depth 1. Second, we develop MILP-optimizations that significantly speed up sensitivity verification for single ensembles and for the first time can also handle multiclass tree ensembles. Third, we introduce a data-aware framework generating realistic examples close to the training distribution. Finally, we conduct an extensive experimental evaluation on large tree ensembles, demonstrating scalability to ensembles with up to 800 trees of depth 8, achieving substantial improvements over the state of the art. This framework provides a practical foundation for analyzing the reliability and fairness of tree-based models in high-stakes applications.