State-of-the-art learning algorithms, such as random forests or neural networks, are often qualified as "black-boxes" because of the high number and complexity of operations involved in their prediction mechanism. This lack of interpretability is a strong limitation for applications involving critical decisions, typically the analysis of production processes in the manufacturing industry. In such critical contexts, models have to be interpretable, i.e., simple, stable, and predictive. To address this issue, we design SIRUS (Stable and Interpretable RUle Set), a new classification algorithm based on random forests, which takes the form of a short list of rules. While simple models are usually unstable with respect to data perturbation, SIRUS achieves a remarkable stability improvement over cutting-edge methods. Furthermore, SIRUS inherits a predictive accuracy close to random forests, combined with the simplicity of decision trees. These properties are assessed both from a theoretical and empirical point of view, through extensive numerical experiments based on our R/C++ software implementation sirus available from CRAN.

Summary Variable importance measures are the main tools used to analyse the black-box mechanisms of random forests. Although the mean decrease accuracy is widely accepted as the most efficient variable importance measure for random forests, little is known about its statistical properties. In fact, the definition of mean decrease accuracy varies across the main random forest software. In this article, our objective is to rigorously analyse the behaviour of the main mean decrease accuracy implementations. Consequently, we mathematically formalize the various implemented MDA algorithms, and then establish their limits when the sample size increases. This asymptotic analysis reveals that these mean decrease accuracy versions differ as importance measures, since they converge towards different quantities. More importantly, we break down these limits into three components: the first two terms are related to Sobol indices, which are well-defined measures of a covariate contribution to the response variance, widely used in the sensitivity analysis field, as opposed to the third term, whose value increases with dependence within covariates. Thus, we theoretically demonstrate that the mean decrease accuracy does not target the right quantity to detect influential covariates in a dependent setting, a fact that has already been noticed experimentally. To address this issue, we define a new importance measure for random forests, the Sobol-mean decrease accuracy, which fixes the flaws of the original mean decrease accuracy, and consistently estimates the accuracy decrease of the forest retrained without a given covariate, but with an efficient computational cost. The Sobol-mean decrease accuracy empirically outperforms its competitors on both simulated and real data for variable selection. An open source implementation in R and C ++ is available online.

Variable importance measures are the main tools to analyze the black-box mechanism of random forests. Although the Mean Decrease Accuracy (MDA) is widely accepted as the most efficient variable importance measure for random forests, little is known about its theoretical properties. In fact, the exact MDA definition varies across the main random forest software. In this article, our objective is to rigorously analyze the behavior of the main MDA implementations. Consequently, we mathematically formalize the various implemented MDA algorithms, and then establish their limits when the sample size increases. In particular, we break down these limits in three components: the first two are related to Sobol indices, which are well-defined measures of a variable contribution to the output variance, widely used in the sensitivity analysis field, as opposed to the third term, whose value increases with dependence within input variables. Thus, we theoretically demonstrate that the MDA does not target the right quantity when inputs are dependent, a fact that has already been noticed experimentally. To address this issue, we define a new importance measure for random forests, the Sobol-MDA, which fixes the flaws of the original MDA. We prove the consistency of the Sobol-MDA and show its good empirical performance through experiments on both simulated and real data. An open source implementation in R and C++ is available online.

Interpretability of learning algorithms is crucial for applications involving critical decisions, and variable importance is one of the main interpretation tools. Shapley effects are now widely used to interpret both tree ensembles and neural networks, as they can efficiently handle dependence and interactions in the data, as opposed to most other variable importance measures. However, estimating Shapley effects is a challenging task, because of the computational complexity and the conditional expectation estimates. Accordingly, existing Shapley algorithms have flaws: a costly running time, or a bias when input variables are dependent. Therefore, we introduce SHAFF, SHApley eFfects via random Forests, a fast and accurate Shapley effect estimate, even when input variables are dependent. We show SHAFF efficiency through both a theoretical analysis of its consistency, and the practical performance improvements over competitors with extensive experiments. An implementation of SHAFF in C++ and R is available online.Preprint. Under review.

No abstract