Interpretable Predictions of Tree-based Ensembles via Actionable Feature Tweaking

By Gabriele Tolomei, Fabrizio Silvestri, Andrew Haines, Mounia Lalmas. 2017

236 citations recorded by the source corpus.

Machine-learned models are often described as "black boxes". In many real-world applications however, models may have to sacrifice predictive power in favour of human-interpretability. When this is the case, feature engineering becomes a crucial task, which requires significant and time-consuming human effort. Whilst some features are inherently static, representing properties that cannot be influenced (e.g., the age of an individual), others capture characteristics that could be adjusted (e.g., the daily amount of carbohydrates taken). Nonetheless, once a model is learned from the data, each prediction it makes on new instances is irreversible - assuming every instance to be a static point located in the chosen feature space. There are many circumstances however where it is important to understand (i) why a model outputs a certain prediction on a given instance, (ii) which adjustable features of that instance should be modified, and finally (iii) how to alter such a prediction when the mutated instance is input back to the model. In this paper, we present a technique that exploits the internals of a tree-based ensemble classifier to offer recommendations for transforming true negative instances into positively predicted ones. We demonstrate the validity of our approach using an online advertising application. First, we design a Random Forest classifier that effectively separates between two types of ads: low (negative) and high (positive) quality ads (instances). Then, we introduce an algorithm that provides recommendations that aim to transform a low quality ad (negative instance) into a high quality one (positive instance). Finally, we evaluate our approach on a subset of the active inventory of a large ad network, Yahoo Gemini.

Subfields: Artificial Intelligence, Safety, Robustness & Interpretability, Signal Processing

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

In Signal Processing

Citations in the corpus

Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AI (2019, neutral): “Finally, a crosswise technique among post-hoc explainability, [240] proposes a framework that poses recommendations that, if taken, would convert an example from one class to another.”
Explaining Predictions from Tree-based Boosting Ensembles (2019, methodology): “(1) We rst extend [13] to work for adaptive boosting [6], which still trains on the original labels.”
A Survey of Methods for Explaining Black Box Models (2018, neutral): “In Reference [121], a solution for the model explanation problem is presented.”
A survey of algorithmic recourse: definitions, formulations, solutions, and prospects (2020, background): “o. diver. spar. ! p opt. cov. rtm. (2014.03) SEDC [119] E heuristic query (2015.08) OAE [48] E ILP white-box (2016.05) HCLS [101, 103] E grad opt/heuristic gradient/query (2017.06) Feature Tweaking [169] E heuristic white-box (2017.11) CF Expl. [177] E grad opt gradient (2017.12) Growing Spheres [1”
Principles and Practice of Explainable Machine Learning (2020, background): “• A different approach on measuring a feature’s importance can be found in [81].”
Questioning the AI: Informing Design Practices for Explainable AI User Experiences (2020, methodology): “rithm Examples Question Type Explain the model Global feature importance Describe the weights of the features used by the model (including visualization that shows the weights of high-level features) [40,59,68, 90] How (Global) Decision tree approximation Approximate the model to an interpretable de”
Interpretable Machine Learning - A Brief History, State-of-the-Art and Challenges (2020, methodology): “Popular local IML methods are Shapley values [69,112] and counterfactual explanations [122,20,81,116,118].”
Explainability fact sheets: a framework for systematic assessment of explainable approaches (2019, background): “When an explanation is provided to help users understand a reason behind a decision, then the users prefer explanations that they can treat as guidelines towards the desired outcome [25, 51].”
Explainable Artificial Intelligence: a Systematic Review (2020, methodology): “[290] introduced an algorithm to tweak the input features to change the output of a tree ensemble classifier.”
Multi-Objective Counterfactual Explanations (2020, methodology): “For the ﬁrst part, we compared MOC – once with and once without the proposed initialization and mutation strategies – with ‘DiCE’ by Mothilal et al. [21], ‘Recourse’ by Ustun et al. [30] and ‘Tweaking’ by Tolomei et al. [29].”
A Survey on the Explainability of Supervised Machine Learning (2020, methodology): “ion Classication Model-specic Non-linear Model Barbella et al. (2009) Cloaking Classication Model-agnostic - Chen et al. (2015) Tweaking Recommendation Classication Model-specic Non-linear Model Tolomei et al. (2017) Nearest Neighbor Classication Model-agnostic - Freitas (2014) SVM Recommendations C”
Interpretable Neural Networks for Predicting Mortality Risk using Multi-modal Electronic Health Records (2019, unknown): “…to support clinical interpretability include building a single tree with multivariate decision nodes [7], extracting an optimal tree with a minimum performance cost [8], and an indirect method that offers recommendations for transforming true negative instances into positively predicted ones [9].”
FOCUS: Flexible Optimizable Counterfactual Explanations for Tree Ensembles (2019, methodology): “(3) This definition aligns with previous ML work on counterfactual explanations [18, 22, 31].”
Counterfactual Explanations of Machine Learning Predictions: Opportunities and Challenges for AI Safety (2019, background): “Tolomei et al. (2017) also use a particular type of counterfactual statements – involving features, which when tweaked transform a true negative instance into a one that a model classifies as positive – generated for tree ensembles and used for improving on-line advertisements.”