Tree Space Prototypes: Another Look at Making Tree Ensembles Interpretable

By Sarah Tan, Matvey Soloviev, Giles Hooker, Martin T. Wells. 2016

79 citations recorded by the source corpus.

Ensembles of decision trees perform well on many problems, but are not interpretable. In contrast to existing approaches in interpretability that focus on explaining relationships between features and predictions, we propose an alternative approach to interpret tree ensemble classifiers by surfacing representative points for each class -- prototypes. We introduce a new distance for Gradient Boosted Tree models, and propose new, adaptive prototype selection methods with theoretical guarantees, with the flexibility to choose a different number of prototypes in each class. We demonstrate our methods on random forests and gradient boosted trees, showing that the prototypes can perform as well as or even better than the original tree ensemble when used as a nearest-prototype classifier. In a user study, humans were better at predicting the output of a tree ensemble classifier when using prototypes than when using Shapley values, a popular feature attribution method. Hence, prototypes present a viable alternative to feature-based explanations for tree ensembles.

Subfields: Artificial Intelligence, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Safety, Robustness & Interpretability

Citations in the corpus

Local Rule-Based Explanations of Black Box Decision Systems (2018, methodology): “Decision trees have been adopted to globally explain neural networks [9, 22] and tree ensembles [18, 34].”
A Survey of Methods for Explaining Black Box Models (2018, background): “The authors of Reference [117] propose the Tree Space Prototype (TSP) f for interpretating tree ensembles b by finding tree prototypes (the comprehensible global predictor cд) in the tree space.”
Meaningful Explanations of Black Box AI Decision Systems (2019, methodology): “For example, decision trees have been adopted to explain neural networks (Krishnan, Sivakumar, and Bhat-tacharya 1999) and tree ensembles (Tan, Hooker, and Wells 2016), while decision rules have been widely used to explain neural networks (Augasta and Kathirvalavakumar 2012; Andrews, Diederich, and…”
Trustworthy AI: From Principles to Practices (2021, methodology): “been proposed to approximate ML models, such as random forests [317, 392] and neural networks [28, 86, 393].”
Born-Again Tree Ensembles (2020, methodology): “Recent works along this line used tree space prototypes to sparsen a tree (Tan et al., 2016) or rectiﬁed decision trees that use hard and soft labels (Bai et al., 2019).”
Principles and Practice of Explainable Machine Learning (2020, background): “The authors in [78] examine ways to inspect the training dataset in order to sample a number of data points that are representative members of their corresponding class.”
Explainable Artificial Intelligence: a Systematic Review (2020, background): “Tree Space Prototypes (TSP) [298] selects prototypes from a training dataset to explain the prediction made be ensembles of DTs and gradient boosted tree models on a new observation.”
Benchmarking and survey of explanation methods for black box models (2021, background): “…PH G/L A PR ps Bien and Tibshirani (2011) TAB IN G/L S mmd- critic Kim et al. (2016) ANY IN G S protodash Gurumoorthy et al. (2019) ANY IN G A tsp Tan et al. (2020) TAB PH L S CF cem Dhurandhar et al. (2018) ANY PH L S cfx Albini et al. (2020) TAB PH L S dice Mothilal et al. (2020) TAB PH L A c-…”
iForest: Interpreting Random Forests via Visual Analytics (2019, methodology): “For example, the similarity metric for random forests can be calculated not only based on feature value distance, but also on how many common leaf nodes are reached [13, 51].”